Tech Ingester

2020-03-16 Tech Ingester: #NYCShutdown

Tech Ingester

Best
New York City to Close Schools, Restaurants and Bars: Live Updates (nytimes.com)
“Mayor Bill de Blasio announced on Sunday night that he would order all bars and restaurants in New York City to close. Restaurants would be limited to takeout and food delivery. Schools, restaurants and bars in New York City will close. Restaurants and bars will be closed in New York City. New York courts postpone many cases. Cuomo asks Trump for military to help fight the pandemic. The number of confirmed cases in New York is now over 700. New York’s presidential primary could be delayed.”

Coronavirus: Why You Must Act Now (medium.com)
This post has received a lot of traction - and criticism due to the author’s lack of credentials - so take with a grain of salt. “It’s a matter of days. Maybe a week or two. When it does, your healthcare system will be overwhelmed...The only way to prevent this is social distancing today. Not tomorrow. Today. That means keeping as many people home as possible, starting now.”

What you can’t say (paulgraham.com)
“Do you have any opinions that you would be reluctant to express in front of a group of your peers? If the answer is no, you might want to stop and think about that.”

Industry
The New Business of AI (and How It’s Different From Traditional Software) (a16z.com)
“In particular, many AI companies have: Lower gross margins due to heavy cloud infrastructure usage and ongoing human support; Scaling challenges due to the thorny problem of edge cases; Weaker defensive moats due to the commoditization of AI models and challenges with data network effects.”

Suckers List: How Allstate’s Secret Auto Insurance Algorithm Squeezes Big Spenders (themarkup.org) - submitted by Alex V.
“Is it O.K. to charge more to that person because they don’t have the resources or time or attention to invest in shopping?... It’s a huge problem that I don’t think society has thought through.”

Startup Option Value Calculator (valuation.vc)
“This calculator was developed by Will Gornall at UBC and Ilya Strebulaev at Stanford University using the methodology developed for their paper Squaring Venture Capital Valuations with Reality.”

Lattice Announces the Invest In Your People Fund (lattice.com) - submitted by Alex V.
“If you’ve worked at Lattice for at least 3 years and part on amicable terms and if you start a new company within 12 months of leaving, Lattice will offer to invest up to $100,000 either at a $5 million valuation or at the terms of a seed round you are raising, whichever is higher. Lattice will look to support these companies through sharing advice, our network, and even partnership where applicable.”

Software
Avoid rewriting a legacy system from scratch, by strangling it (understandlegacycode.com)
“Progressively delete the old code base, in favor of a new one. Have the new code acts as a proxy for the old code. Users use the new system, but it just redirects to the old one. Re-implement each behavior to the new codebase, with no change from the end-user perspective. Progressively fade away the old code by making users consume the new behavior. Delete the old, unused code.”

Misc
Money Stuff - Bloomberg Opinion (link.mail.bloombergbusiness.com)
“A daily take on Wall Street, finance, companies and other stuff.” - from Matt Levine. Lots of amusing insight.

Software Lead Weekly (softwareleadweekly.com)
A weekly email for busy people who care about people, culture and leadership.

The Mandalorian: This Is the Way (ascmag.com)
“Cinematographers Greig Fraser, ASC, ACS and Barry “Baz” Idoine and showrunner Jon Favreau employ new technologies to frame the Disney Plus Star Wars series.”

Why you should define your fears instead of your goals | Tim Ferriss (youtube.com)
A 3-page process for overcoming your fears preventing you from doing X. Page 1: If I do X, 1) define fears, 2) what you can do to prevent them, 3) how you could repair damage if fears come true. Page 2: What are the benefits of an attempt or partial success? Page 3: The cost of inaction (emotional/physical/financial) over 6 months, 1 year, 3 years.

New Atheism: The Godlessness that failed (slatestarcodex.com)
“The concept of the Internet as magical place where you could change other people’s minds had given way to the Internet as magical place where you could complain to like-minded friends about how ignorant other people were.”

Have an interesting article to share? Send it to me and get featured!
Want to get in touch? Reply here, or visit nicholasbradford.io
Haven’t signed up yet? Click here.
Views expressed here are my own and not representative of any past or present affiliated entity.

2020-02-17 Tech Ingester: Fat Tails, Exploding Topics, and Intentional Latency

Tech Ingester

Best
How to Get Lucky: Focus On The Fat Tails (taylorpearson.me)
A very practical intro to “80/20 Thinking”, a more positive lens through which to view the “Black Swan” idea. “Let’s say you have $500 to spend to help grow your business… would you spend it on pay-per-click ads driving everyone to your products, or buying a group of people in your industry dinner? Ads is bell curve thinking, hosting dinner is 80/20 thinking.”

Intro to Program Synthesis (csail.mit.edu)
Free materials for a class on Program Synthesis, “a new field at the intersection of programming languages, formal methods and AI.” I’m starting this week with some friends, let me know if you’re interested in the study group!

Industry
Discover Exploding Topics (explodingtopics.com)
“We surface rapidly growing topics before they take off.”

Software
I Add 3-25 Seconds of Latency to Every Page I Visit (howonlee.github.io)
“...if you can inject latency into sites artificially, you can reduce the actual impact of the addiction in a controllable way while not denying the enjoyment of the Internet to yourself.”

10 tips for reviewing code you don’t like (developers.redhat.com)
“Avoid hyperbolic or bombastic assertions, avoid argument strategies, avoid elitist or demeaning language, and avoid constructs like “obviously” and “why don’t you just…”. Use clear, factual statements and supportive language, ask questions, and move things forward. Remember that coworkers and contributors are human people, and their time is worthy of the same respect as yours.”

Awesome functional python (github.com)
“A curated list of awesome things related to functional programming in Python.”

Someone Used Neural Networks To Upscale An 1895 Film To 4K 60 FPS (digg.com)
Wow :O

Misc
Book Review: Against the Grain (slatestarcodex.com)
“You can’t tax hunter-gatherers, because you don’t know how many they are or where they are… you also can’t tax potato farmers, because they can just leave when they hear you coming, and you will never be able to find all of the potatoes and dig them up and tax them… But you can tax grain farmers!” Not unrelated to “Sapiens.”

How to Feel Nothing Now, in Order to Feel More Later (nytimes.com)
They would not be eating. They would not look at any screens. They would not listen to music. They would not exercise. They would not touch other bodies for any reason… The number of things to not do is potentially endless.

Tying cables (youtube.com)
Never get your cables tangled again! (assuming you’re not using AirPod Pros)

2019-12-30 Tech Ingester: Career Advice, Hiring the Trendiest, and Genius

Tech Ingester

Best
Don't Call Yourself A Programmer, And Other Career Advice (kalzumeus.com)
“Engineers are hired to create business value, not to program things… Producing beautiful software is not a goal. Solving complex technical problems is not a goal. Writing bug-free code is not a goal. Using sexy programming languages is not a goal. Add revenue. Reduce costs. Those are your only goals.”

Industry
The Lesson to Unlearn (paulgraham.com)
“The most damaging thing you learned in school wasn't something you learned in any specific class. It was learning to get good grades… Why did founders tie themselves in knots doing the wrong things when the answer was right in front of them? Because that was what they'd been trained to do. Their education had taught them that the way to win was to hack the test.”

We only hire the trendiest (danluu.com)
“...mostly due to superficial factors that don't have much to do with actual productivity. This is a really common story among people who end up at Google. If you hired them before they worked at Google, you might have gotten a great deal! But no one (except Google) was willing to take that chance.”

Software
What nobody tells you about documentation (divio.com)
“They are: tutorials, how-to guides, explanation and technical reference. They represent four different purposes or functions, and require four different approaches to their creation. Understanding the implications of this will help improve most software documentation - often immensely.”

How to do a code review (google.github.io)
From Google's Engineering Practices documentation. “The Standard of Code Review, What to Look For In a Code Review, Navigating a CL in Review, Speed of Code Reviews, How to Write Code Review Comments, Handling Pushback in Code Reviews.”

Misc
Advice (patrickcollison.com)
From Founder/CEO of Stripe: “Go deep on multiple things. Don't make the mistake of judging your success based on your current peer group. Make things. Operating in a space with a lot of uncertainty is a very different experience to learning something. Heuristic: do your friends at school think your path is a bit strange? If not, maybe it's too normal.”

The Bus Ticket Theory of Genius (paulgraham.com)
“To do great work you need both natural ability and determination. But there's a third ingredient that's not as well understood: an obsessive interest in a particular topic… If I had to put the recipe for genius into one sentence, that might be it: to have a disinterested obsession with something that matters.”

2019 Adversarial Collaboration Entries (slatestarcodex.com)
Where two people with opposite views on a controversial issue work together to present a unified summary of the evidence and its implications. Including: “Does calorie restriction slow aging?”, “Should we colonize space to mitigate x-risk?”, and “Should gene editing technologies be used in humans”.

Financial Incentives Are Weaker Than Social Incentives But Very Important Anyway (slatestarcodex.com)
“The article starts with some surprising facts. Increased taxes on the rich don’t make rich people work much less. Salary caps on athletes don’t decrease athletic performance. Increased welfare doesn’t make poor people work less. Decreased job opportunities in one area rarely cause people to move elsewhere… If it is not financial incentives, what else might people care about? The answer is something we know in our guts: status, dignity, social connections.”

Having Kids (paulgraham.com)
“When people had babies, I congratulated them enthusiastically, because that seemed to be what one did. But I didn't feel it at all. "Better you than me," I was thinking...I had plenty of happy times before I had kids. But if I count up happy moments, not just potential happiness but actual happy moments, there are more after kids than before. Now I practically have it on tap, almost any bedtime.”

Tech Ingester: Counterfeit Capitalism, Liberating Constraints, and Dark Money in Almonds

Tech Ingester

Best
Too Much Dark Money in Almonds (slatestarcodex.com)
“Why is there so little money in politics? ...Add up all US spending on candidates, PACs, lobbying, think tanks, and advocacy organizations – liberal and conservative combined – and we’re still $2 billion short of what we spend on almonds each year. In fact, we’re still less than Elon Musk’s personal fortune; Musk could personally fund the entire US political ecosystem on both sides for a whole two-year election cycle.”

Industry
WeWork and Counterfeit Capitalism (substack.com)
“The goal of Son, and increasingly most large financiers in private equity and venture capital, is to find big markets and then dump capital into one player in such a market who can underprice until he becomes the dominant remaining actor...This kind of counterfeit capitalism is terrible for society as a whole.”

Software
Software Architecture is Overrated, Clear and Simple Design is Underrated (pragmaticengineer.com)
“Let me start with a few things that might sound surprising. First, none of these designs used any of the standard software architecture planning tools. We did not use UML, nor the 4+1 model, nor ADR, nor C4, nor dependency diagrams. We created plenty of diagrams, but none of them followed any strict rules. Just plain old boxes and arrows.”

Constraints Liberate, Liberties Constrain — Runar Bjarnason (youtube.com)
“As programmers, we tend to think of expressive power of a language or library as an unmitigated good. In this talk I want to show the contrary; that restraint and precision are usually better than power and flexibility. A constraint on component design leads to freedom and power when putting those components together into systems.”

Reverse interview (github.com)
“This is a list of questions which may be interesting to a tech job applicant. The points are not ordered and many may not apply to a given position, or work type. It was started as my personal list of questions, which grew over time to include both things I'd like to see more of and red flags which I'd like to avoid.”

Misc
The Hedge Fund Billionaire’s Guide to Buying Your Kids a Better Shot at Not Just One Elite College, but Lots of Them (propublica.org)
“Less well publicized, however, are the various ways in which Shaw has applied his fund’s risk-averse, quantitative approach to nearly every aspect of his life... It was company lore that before Shaw traveled, an assistant would take the exact same trip — same car service, same airport, same seat on the plane — to eliminate any inefficiencies.”

Scott’s Supreme Quantum Supremacy FAQ! (scottaaronson.com)
“A group at Google has now achieved quantum computational supremacy with a 53-qubit superconducting device...because so much misinformation is swirling around, what I thought I’d do here, in my role as blogger and “public intellectual,” is offer Scott’s Supreme Quantum Supremacy FAQ.”

2019-09-09 Tech Ingester: Boring Technology, Consumption/Creation, and Growing a Language

Tech Ingester

Best
Choose Boring Technology (boringtechnology.club)
“I gave this most recently at the WikiMedia Foundation’s developer conference, where Scott Ananian called it “how to be old, for young people.””

Consume less, create more (tjcx.me)
“Like most people, I like to sit on the bus, stare out at the zombies drooling over their phones, and feel smugly superior. I see their bored, lifeless eyes and pity these sheeple whose lives have been ruined in service of Big Tech profits. This lasts for about thirty seconds before I get bored.”

Software
Growing a Language, by Guy Steele (youtube.com)
This talk comes with a very interesting meta-format. “Guy Steele's keynote at the 1998 ACM OOPSLA conference on "Growing a Language" discusses the importance of and issues associated with designing a programming language that can be grown by its users.”

John Hughes - Don't Write Tests (youtube.com)
“Writing test cases is laborious, often a little neglected, and ultimately rather unsatisfying, since errors slip through even the most comprehensive test suites. Can we generate tests instead? In this talk I’ll show test case generation using QuickCheck, drawing on experience of applying it at scale at Quviq. I’ll try to explain why property-based testing can be so effective, and how to use it most effectively.”

"Haxl: A Big Hammer for Concurrency" by Simon Marlow (youtube.com)
“Our programming languages are usually "sequential by default", and you have to be explicit if you want concurrency. Much of the code we write, however, is insensitive to reordering, and in these cases concurrency is actually a better default. In this talk I'll describe how the Haxl framework developed and open-sourced by Facebook flips the default from sequential to concurrent. I'll show how that can be useful, and give some insight into how it works.”

Industry
All about Direct Listings (a16z.com)
“in a Direct Listing, no shares are sold by the company itself, and therefore no capital is raised. So why would a company do a Direct Listing? What are other key differences between an IPO and a Direct Listing? And what are the tradeoffs involved in this new approach? Since the Direct Listings process is so new, I try to demystify the topic in this post given my own vantage point immersed in the capital markets (including behind the scenes with Slack, where we’re investors)."

Tech Interview Handbook (github.com)
Most comprehensive I’ve seen so far. “Carefully curated content to help you ace your next technical interview”

Square’s Growth Framework for Engineers and Engineering Managers (squareup.com)
“Our Software Engineering Career Ladder is a crucial element of Square’s toolkit for supporting engineers’ growth. It helps us ensure that we evaluate and reward engineers consistently and fairly, regardless of their team, discipline, or background. Every employee has a level, which describes the scope, complexity, and impact of their role and factors into compensation.”

Productivity
Every productivity thought I've ever had, as concisely as possible (guzey.com)
With specific tips on pomodoro variations, changing locations, and most importantly: “Every productivity system stops working eventually and there’s nothing you can do about it.”

Misc
How One Guy Sold the World on an $80 Used Tissue (youtube.com)
“Using only $1,000 over the course of one week, Mekki Leeper sets out to fabricate an absurd alternative medicine company and fool the world into thinking it’s real.”

How Tax Policy Gave Us White Claw (nymag.com)
“Because White Claw is brewed like beer, it’s taxed like beer, which is important because beer is taxed in the U.S. at a much lower rate than spirits. If you made a product similar to White Claw by mixing vodka with seltzer and putting it in a can, a six-pack would be subject to almost $2 in additional taxes when sold in New York City.”

The Museum of Future Experiences offers a spooky, surreal take on VR (techcrunch.com)
I personally visited their NYC pop-up recently - though I can’t say I enjoyed it as much as TechCrunch did.

2019-07-01 Tech Ingester: PR Genius, Epistemic Helplessness, and Social Status

Tech Ingester

Best
Series on Epistemic Learned Helplessness (slatestarcodex.com)
Posts on the topic: “What if there are some issues where rational debate inherently leads you astray?”

Industry
Mary Meeker’s most important trends on the internet (vox.com)
“It’s the holiday season for data nerds: That is, Mary Meeker is delivering her annual Internet Trends Report — the most highly anticipated slide deck in Silicon Valley — again at Code Conference 2019.”

Amazon’s PR Genius (ben-evans.com)
“Nothing it cares about ever leaks. Almost all of the press coverage, even the negative stories, runs to a script that Bezos could have written - "We do amazing things to get low prices to customers" and "it's incredibly hard to compete with us".”

Startup idea checklist (defmacro.org)
“I use this list both to develop ideas and filter them. If you adopt it, be careful about using it as a filter. Remember that in the early stages, good ideas are very easy to kill.”

Software
Release Scala 2.13.0 (github.com)
“Standard library collections have been overhauled for simplicity, performance, and safety. This is the centerpiece of the release...No more CanBuildFrom. Transformation methods no longer take an implicit CanBuildFrom parameter.” For reference on why this is a big deal, see “Is the Scala 2.8 collections library a case of “the longest suicide note in history?”

Things I Learnt The Hard Way (in 30 Years of Software Development) (juliobiason.net)
“Again, some things are really cynical, others are long observations on different jobs.”

The Data Engineering Cookbook (github.com)
A very comprehensive guide on the skills and technologies you’ll need, with plenty of case studies for the advanced reader.

AI/Machine Learning
Towards federated learning at scale: system design (acolyer.org)
“Here we’re actually training the model in a distributed fashion, using data collected on the devices, without the data ever leaving those devices.”

Misc
Social Status: Down the Rabbit Hole (meltingasphalt.com)
“The Golden Rule of politics, which has arguably created more prosperity for our species [than the golden rule of ethics], goes something like this: “Admire those who would make good allies.” Unlike the ethical rule… the political rule comes to us quite naturally. We're sycophants, one and all.”

If Kim Jong-Un Opened A KFC, Would You Eat There? (slatestarcodex.com)
“Tobacco-free cigarettes have helped a lot of people quit smoking; meat substitutes have helped a lot of people (recently sort of including me) become vegetarian. I want a smoke-free meatless future. But does it become a mockery when the same companies that provided the smoky meaty past are selling it to us? If they make a fortune being evil, resist change, and lose, should they get to make a second fortune being good?”

Trolley: A community for new grads in New York City. (trolley.to)
The pitch is to be able to connect with other new grads, receive “advice on #adulting”, and more. Will keep you posted on the results.

2019-05-20 Tech Ingester: Monetary Policy, Pitch Rooms, and Applied Math

Tech Ingester

Industry
A Tax That Could Fix Big Tech (nytimes.com)
Putting a levy on targeted ad revenue would give Facebook and Google a real incentive to change their dangerous business models.

The a16z Pitch Room: Sandbox VR (a16z.com)
“Have you ever wished you could be in the room when founders present to the Andreessen Horowitz investing team? Even better, do you wish you could stay to hear the discussion after the pitch? We hear you and are thrilled to launch our latest series of videos called “The a16z Pitch Room”. This inaugural episode features the Sandbox VR team Steve Zhao and Siqi Chen pitching to General Partner Andrew Chen.”

Software
What is Clean Code? (informit.com)
Excerpt from the beginning of “Clean Code” by Uncle Bob Martin: The “grand redesign” never happens. Clean code: is elegant, does one thing well, is efficient (a form of elegance), is readable, and concise (including in number of entities). Bad code: tempts you to extend it (“broken windows”), is surprising, and makes the language look ill-suited for the task.

Category Theory for Programmers (bartoszmilewski.com)
“Composition is at the very root of category theory — it’s part of the definition of the category itself. And I will argue strongly that composition is the essence of programming. We’ve been composing things forever, long before some great engineer came up with the idea of a subroutine.”

Misc
It’s Time to Look More Carefully at “Monetary Policy 3 (MP3)” and “Modern Monetary Theory (MMT)” (linkedin.com)
“[Monetary policy] won’t work hardly at all in stimulating economic prosperity in the ways that we are used to having it stimulate economic activity, which are through interest rate cuts (what I call Monetary Policy 1) and through quantitative easing (what I call Monetary Policy 2). That is because it won’t be effective in producing money and credit growth (i.e., spending power) and it won’t be effective in getting it in the hands of most people to increase their productivity and prosperity.”

Self Studying the MIT Applied Math Curriculum (harshsikka.me)
“I've decided to engage in a pretty interesting self development project this summer: Completing the requirements for the MIT Applied Math curriculum through auditing classes, meeting professors, and completing the corresponding OCW course assignments and exams where applicable.”

2019-04-22 Tech Ingester: Late software projects, startup stock options, and death by powerpoint

Tech Ingester

Best
Programming: doing it more vs doing it better (kevinmartinjose.com)
Quantity or quality? One story: “a curious fact emerged: the works of highest quality were all produced by the group being graded for quantity. It seems that while the “quantity” group was busily churning out piles of work – and learning from their mistakes – the “quality” group had sat theorizing about perfection, and in the end had little more to show for their efforts than grandiose theories.” Another: “a writer is someone for whom writing is more difficult than it is for other people. The best writers write much more slowly than everyone else, and the better they are, the slower they write.”

Industry
Startup Stock Options – Why A Good Deal Has Gone Bad (steveblank.com)
“VC’s have just changed the ~50-year old social contract with startup employees. In doing so they may have removed one of the key incentives that made startups different from working in a large company.”

Indeed, it seems that Google IS forgetting the old Web (zona-m.net)
“I think Google has stopped indexing the older parts of the Web. I think I can prove it. Google’s competition is doing better.”

Software
Why software projects take longer than you think – a statistical model (erikbern.com) - from Alex V.
“If my model is right...: People estimate the median completion time well, but not the mean. The mean turns out to be substantially worse than the median, due to the distribution being skewed (log-normally). When you add up the estimates for n tasks, things get even worse. Tasks with the most uncertainty (rather the biggest size) can often dominate the mean time it takes to complete all tasks.”

Tweets by @hillelogram (twitter.com) - from Noah S.
“One of my most controversial software opinions is that your sleep quality and stress level matter far, far more than the languages you use or the practices you follow. Nothing else comes close: not type systems, not TDD, not formal methods, not ANYTHING. Allow me to explain why.”

You Are Not Google (bradfieldcs.com)
“The thing is there’s like 5 companies in the world that run jobs that big. For everybody else… you’re doing all this I/O for fault tolerance that you didn’t really need. People got kinda Google mania in the 2000s: “we’ll do everything the way Google does because we also run the world’s largest internet data service””

Misc
Death by PowerPoint: the slide that killed seven people (mcdreeamiemusings.com)
“Finally the single most important fact, that the foam strike had occurred at forces massively out of test conditions, is hidden at the very bottom. Twelve little words which the audience would have had to wade through more than 100 to get to. If they even managed to keep reading to that point. In the middle it does say that it is possible for the foam to damage the tile. This is in the smallest font, lost.”

2019-03-25 Tech Ingester: 737 Max, Underrated Applicatives, and the Death of the Newsfeed

Tech Ingester

Best
How a 50-year-old design came back to haunt Boeing with its troubled 737 Max jet (latimes.com) - sourced by Ollie C.
“To handle a longer fuselage and more passengers, Boeing added larger, more powerful engines, but that required it to reposition them to maintain ground clearance. As a result, the 737 can pitch up under certain circumstances. Software... was added to counteract that tendency… The software erroneously thought the aircraft was at risk of losing lift and stalling — because of a malfunctioning sensor — and ordered the stabilizer at the rear to put it into a series of sharp dives that ultimately caused the plane to crash into the Java Sea.”

Industry
The Death of the Newsfeed (ben-evans.com)
“50% of Facebook's engineering effort goes into stuffing more noise into the newsfeed, and the other 50% into working out ways to filter it.”

Software
The underrated applicative functor (softwaremill.com)
“The “scream” operator |@| finally gets us what we wanted - our independent computations are now represented in a much elegant way. The monadic flow may look like imperative code, while here we have a more functional representation of our actual intent. Apart from adjusting code structure we help future readers to recognise parts that can be parallelized or refactored due to their independence.” (a little out of date - the terrible |@| syntax is no longer necessary)

Why is it faster to process a sorted array than an unsorted array? (stackoverflow.com)
You are a processor and you see a branch. You have no idea which way it will go. What do you do? You halt execution and wait until the previous instructions are complete. Then you continue down the correct path. Modern processors are complicated and have long pipelines. So they take forever to "warm up" and "slow down". Is there a better way? You guess which direction the branch will go! See this blog on branch prediction for more: https://danluu.com/branch-prediction/.

Why is this program erroneously rejected by three C++ compilers? (stackoverflow.com)
Ever tried drawing out your code by hand?

Misc
What the Hell is Going on? (perell.com) - sourced by Brendan M.
“The same shift from information scarcity to information abundance is reshaping commerce, education, and politics...By changing the balance of informational influence, the internet inverts the Mass Media paradigm.”

Four Steps to an Applied Micro Paper (brown.edu) - sourced by Alex V.
“Write an introduction for the paper you aspire to write. Feel free to make up your results, within reason. When you think you are done, ask yourself: if I write the paper outlined in this introduction, will I be happy with it?”

2019-03-11 Tech Ingester: Controversy, Free vs. Tagless, and Humans as Non-General Intelligences

Tech Ingester

Best
Sort by Controversial (slatestarcodex.com)
“That was 8 PM. We’d been standing in Brad’s office fighting for five hours. At 8:01, after David and Shiri had stormed out, we all looked at each other and thought – holy shit, the controversial filter works.” (fiction)

Software
Famous Laws Of Software Development (timsommer.be)
My favorite is Hofstadter's Law: ‘It always takes longer than you expect, even when you take into account Hofstadter's Law."

Free and tagless compared - how not to commit to a monad too early (softwaremill.com)
“Quite often we see values wrapped in monads in our method signatures: be it Future, Task or DBIOAction. Which monad to choose as "the" monad which we will predominantly use is often an important design decision. Maybe it's a good idea to defer this commitment as long as possible? Free monads and final tagless encoding offer a solution.”

AI/Machine Learning
Humans Who Are Not Concentrating Are Not General Intelligences (wordpress.com)
“The scary thing about GPT-2-generated text is that it flows very naturally if you’re just skimming, reading for writing style and key, evocative words... I would not have noticed they were machine-generated. I would not have noticed anything amiss about them at all. But if I read with focus, I notice that they don’t make a lot of logical sense.”

GAN dissection: visualizing and understanding generative adversarial networks (acolyer.org)
“Turning off (ablating) units identified as associated with common object classes causes the corresponding objects to mostly disappear from the generated scenes...By forcing units on we can try to insert objects into scenes. For example, activating the same ‘door units’ across a variety of scenes causes doors to appear – but the actual appearance of the door will vary in accordance with the surrounding scene.”

2019-02-25 Tech Ingester: Personal Infrastructure, Failing to Build a Billion-Dollar Company, and Secret Histories

Tech Ingester

Best
Seeking the Productive Life: Some Details of My Personal Infrastructure (stephenwolfram.com)
“At an intellectual level, the key to building this infrastructure is to structure, streamline and automate everything as much as possible—while recognizing both what’s realistic with current technology, and what fits with me personally. In many ways, it’s a good, practical exercise in computational thinking.”

Reflecting on My Failure to Build a Billion-Dollar Company (medium.com)
‘In 2011, I left my job as the second employee at Pinterest — before I vested any of my stock — to work on what I thought would be my life’s work. I thought Gumroad would become a billion-dollar company, with hundreds of employees. It would IPO, and I would work on it until I died. Something like that.”

Industry
Programming Interview Questions Are Too Hard and Too Short (triplebyte.com)
“Whether or not a candidate answers a question correctly is not the only source of signal during an interview. You can also evaluate their process by, for example, observing how long it takes them to finish, how clean their code is, and how much they struggle while finding a solution. Our analysis shows that this second source of signal (process) is almost as predictive as the first (correctness).”

How to Choose a Startup to Work For by Thinking Like An Investor (triplebyte.com)
You’re looking for personal networks, learning opportunities and equity. “My advice would be to focus on how impressive the accomplishments of the founders are relative to their peer group and environment... You also need to evaluate how well a founder understands the problem they're working on...Evaluating the relationship between founders is as important as evaluating the founders themselves. The top cause of startup death during a Y Combinator batch was cofounder disputes...If you are an engineer, I'd also ask specific questions around how product ideas are generated and work is assigned.”

The Secret History of Women in Coding (nytimes.com)
“Computer programming once had much better gender balance than it does today. What went wrong?”

AI/Machine Learning
Better Language Models and Their Implications (openai.com)
“We’ve trained a large-scale unsupervised language model which generates coherent paragraphs of text, achieves state-of-the-art performance on many language modeling benchmarks, and performs rudimentary reading comprehension, machine translation, question answering, and summarization — all without task-specific training.” Controversial - the trained model was not released for ethical concerns over its potential misuse.

Misc
Compounding Knowledge (fs.blog)
“A lot of us are on the treadmill of consuming expiring information. Not Buffett. He filled his mental filing cabinet with information that had a long half-life.”

2019-02-04 Tech Ingester: The App Scandal, DeepStar, and Losing Money

Tech Ingester

Industry
Everything you need to know about Facebook, Google’s app scandal (techcrunch.com)
“Facebook and Google landed in hot water with Apple this week after two investigations by TechCrunch revealed the misuse of internal-only certificates — leading to their revocation, which led to a day of downtime at the two tech giants.”

Major iPhone FaceTime bug lets you hear the audio of the person you are calling … before they pick up (9to5mac.com)
“A significant bug has been discovered in FaceTime and is currently spreading virally over social media. The bug lets you call anyone with FaceTime, and immediately hear the audio coming from their phone — before the person on the other end has accepted or rejected the incoming call. Apple says the issue will be addressed in a software update “later this week”.”

How Companies Secretly Boost Their Glassdoor Ratings (wsj.com)
“Employers flood the ranking site with 5-star postings requested from enthusiastic staffers, leading to unusual spikes, a WSJ investigation found.”

Trouble hiring senior engineers? It's probably you (hiringengineersbook.com)
“There are many reasons why some teams fail to attract good talent. But all the teams that have an easy time have realized a simple fact about our current market situation: When hiring senior engineers, the company doesn’t choose the candidate, the candidate chooses the company.”

AI/Machine Learning
AlphaStar: Mastering the Real-Time Strategy Game StarCraft II (deepmind.com)
“StarCraft, considered to be one of the most challenging Real-Time Strategy (RTS) games and one of the longest-played esports of all time, has emerged by consensus as a “grand challenge” for AI research. Now, we introduce our StarCraft II program AlphaStar, the first Artificial Intelligence to defeat a top professional player.” Check out some external analysis (usejournal.com) critiquing AlphaStar’s use of superhuman speed and precision, and videos of the full commentated matches (blizzard.com).

L2 Regularization and Batch Norm (janestreet.com)
“In particular, when used together with batch normalization in a convolutional neural net with typical architectures, an L2 objective penalty no longer has its original regularizing effect. Instead it becomes essentially equivalent to an adaptive adjustment of the learning rate!”

Misc
Why Are Young People Pretending to Love Work? (nytimes.com)
“I saw the greatest minds of my generation log 18-hour days — and then boast about #hustle on Instagram. When did performative workaholism become a lifestyle?”

How To Be Successful (samaltman.com)
“I’ve observed thousands of founders and thought a lot about what it takes to make a huge amount of money or to create something important. Usually, people start off wanting the former and end up wanting the latter. Here are 13 thoughts about how to achieve such outlier success. “

The Miracle Clinic (vice.com)
“It’s a calm scene, and it belies the vital role the Mae Tao Clinc has played for nearly three decades in the middle of the world’s longest-running civil war. Around 350 people arrive her every day seeking all kinds of care - ophthamology, surgery, acupuncture - but maternal health is Mae Tao’s specialty.”

Losing Money Reduces the Risk of Too Much Money (bloomberg.com)
“After seven straight years of poor performance, and the ensuing billions of dollars of redemptions by irate clients, Einhorn took stock of the situation and decided: Hey, good news, we have solved the problem of growing too fast! That risk has been taken right off the table! And that news was so good that he had to share it with clients. And it’s good news for the clients too, because, if they haven’t taken their money out of Greenlight in disappointment after seven straight bad years, now they can put more money in. They can put more money in! What?”

How Blind Contour Drawing Can Help You Become a Better Artist (artsy.net)
“But when done correctly––meaning, without peeking or removing the tip of the pen from the page––blind contour drawing gives artists an opportunity to portray exactly what they observe, instead of what they think their subject is supposed to look like. “A contour drawing is like climbing a mountain as contrasted with flying over it with an airplane,” Nicolaïdes wrote. “It is not a quick glance at the mountain from far away, but a slow, painstaking climb over it, step by step.””

2019-01-22 Tech Ingester: Culture, Money and Happiness, and the American AI Attitude

Tech Ingester

Best
Everything you need to know about whether money makes you happy (80000hours.org)
“It’s a cliché that “you can’t buy happiness”, but at the same time, financial security is among most people’s top career priorities. Moreover, when people are asked what would most improve the quality of their lives, the most common answer is more money. What’s going on here? Who is right?”

Industry
Inside Facebook's 'cult-like' workplace, where dissent is discouraged and employees pretend to be happy all the time (cnbc.com)
“More than a dozen former Facebook employees detailed how the company’s leadership and its performance review system has created a culture where any dissent is discouraged. Employees say Facebook’s stack ranking performance review system drives employees to push out products and features that drive user engagement without fully considering potential long-term negative impacts on user experience or privacy.”

Netflix Culture (netflix.com)
“Our core philosophy is people over process. More specifically, we have great people working together as a dream team.” See the [somewhat famous] slides here.

Paul English of Kayak, on Nurturing New Ideas (nytimes.com)
“When I gave people their performance reviews, I would literally take a crinkled envelope, and I’d write five words on it... I’d say to them, “Let’s say I left this company, and five years from now I was sitting in a bar and someone said, ‘Hey, what’s that guy like?’ ... And there are two or three words that are positive, and there are two or three words that are really negative.” … One guy in particular e-mailed me 10 years later, and claimed that he still carries that piece of paper around.”

Software
Hacker Responsible for 51% Attack Against Ethereum Classic Returns Part of the Stolen Funds (ethereumworldnews.com)
“Based on our analysis, the hashing power of ETC network is still not strong enough, and it’s still possible to rent enough hashing power to launch another 51% attack.”

Why I Can No Longer Recommend Google Fi (onemileatatime.com)
“If a card in your Google Pay is stolen, or someone uses your Payments account fraudulently, or anything happens that leads to a security flag being raised, it can lead to your Google Payments account being frozen… The only suggestion of a solution we’ve been given is that he abandon both his email address and phone number of the past twenty years and start fresh.”

AI/Machine Learning
MIT Deep Learning (mit.edu)
“This page is a collection of MIT courses and lectures on deep learning, deep reinforcement learning, autonomous vehicles, and artificial intelligence taught by Lex Fridman.”

Artificial Intelligence: American Attitudes and Trends (github.io)
“our findings raise more questions than they answer; they are more suggestive than conclusive. Accordingly, we recommend caution in interpreting the results; we confine ourselves to primarily reporting the results...There is substantially more support for developing high-level machine intelligence by those with larger reported household incomes, such as those earning over $100,000 annually (47%) than those earning less than $30,000 (24%); by those with computer science or programming experience (45%) than those without (23%); by men (39%) than women (25%). These differences are not easily explained away by other characteristics (they are robust to our multiple regression).”

Misc
Bridgewater's Pure Alpha Fund Returned 14.6% Last Year (bloomberg.com)
“Hedge funds on average lost 6.7 percent in 2018, according to the HFRX Global Hedge Fund Index. That compares with a 4.4 percent decline for the S&P 500 Index of stocks.” Interestingly, 2018 was the first full year I worked there. Coincidence?

Penguin travels every year to visit man who rescued him (cbc.ca)
“Ever since a 71-year-old Brazilian man rescued a struggling penguin, he's been receiving regular visits from his feathered friend.”

Twelve Things I Never Knew About Clothes Until I Became a Personal Shopper for Barneys
“Robert Nguyen, Barneys’s personal shopping and studio services manager, estimates that the department collectively claims at least 20 clients who will dish out over $1 million for fashion each year. And each of the stylists has a handful of shoppers who will easily drop $250,000 to $300,000 every season. Sometimes customers spend this much in one day: The biggest single sprees on record have totaled more than $400,000.”

2019-01-07 Tech Ingester: Public Domain Day, Coconut, and Web Annotation

Tech Ingester

Industry
January 1, 2019 is (finally) Public Domain Day: Works from 1923 are open to all! (duke.edu)
“Works from 1923 were set to go into the public domain in 1999, after a 75-year copyright term. But in 1998 Congress hit a two-decade pause button and extended their copyright term for 20 years.”

The Yoda of Silicon Valley (nytimes.com)
Donald Knuth, master of algorithms, reflects on 50 years of his opus-in-progress, “The Art of Computer Programming.”

Software
Software Engineering at Google (arxiv.org)
“We catalog and describe Google's key software engineering practices” - granted, in the rosiest possible terms.

Awesome Interviews (github.com)
“A curated list of lists of technical interview questions” for various programming languages, databases, algorithms, design patterns, and more.

Coconut (http://coconut-lang.org/)
“A functional programming language that compiles to Python. Since all valid Python is valid Coconut, using Coconut will only extend and enhance what you're already capable of in Python.” Installation is as easy as `$ pip install coconut`.

Misc
Annotate the web, with anyone, anywhere. (hypothes.is)
“We’re a nonprofit on a mission to bring an open conversation over the whole web. Use Hypothesis right now to hold discussions, read socially, organize your research, and take personal notes.”

IQ is largely a pseudoscientific swindle (medium.com)
By Nassim Nicholas Taleb (writer of “The Black Swan”): ““IQ” is a stale test meant to measure mental capacity but in fact mostly measures extreme unintelligence (learning difficulties), as well as, to a lesser extent, a form of intelligence, stripped of 2nd order effects. It is via negativa not via positiva. Designed for learning disabilities, it ends up selecting exam-takers, paper shufflers, obedient IYIs (intellectuals yet idiots), ill adapted for “real life”.”

Want to get in touch? Reply here, or visit nicholasbradford.io
Haven’t signed up yet? Click here.
Views expressed here are my own and not representative of any past or present affiliated entity.

2018-12-31 Tech Ingester: Productivity, Market Moves, and the Friendship that Made Google Huge

Tech Ingester

Best
Productivity (github.com)
The new and improved productivity advice summary, courtesy of yours truly - with all-new features such as music playlist organization, computer directory structure, and more.

Industry
Pronto Means Ready (medium.com)
By Anthony Levandowski. “I know what some of you might be thinking: “He’s back?” Yes, I’m back.”

Startup Pitch Decks (airtable.com)
A huge collection of pitch decks from various funding stages.

Software
The Friendship That Made Google Huge (newyorker.com)
“Coding together at the same computer, Jeff Dean and Sanjay Ghemawat changed the course of the company—and the Internet.” As somebody who normally dislikes pair programming, I found this article especially interesting.

Guesstimate: A spreadsheet for things that aren’t certain (getguesstimate.com)
“Make a great estimate in seconds. If you think a number is between 5 and 9, simply write "5 to 9". Guesstimate uses Monte Carlo sampling to correctly estimate uncertain results.”

Misc
To Help Put Recent Economic & Market Moves in Perspective (linkedin.com)
“I’d like you to see [recent economic & market moves] within the context of how the market and economic machines work because if you understand that, you can understand these things on your own.”

Want to get in touch? Reply here, or visit nicholasbradford.io
Haven’t signed up yet? Click here.
Views expressed here are my own and not representative of any past or present affiliated entity.

2018-12-24 Tech Ingester: Billboards, Robinhood’s backtracking, and the Efficacy of Parachutes

Tech Ingester

Best
Parachute use to prevent death and major trauma when jumping from aircraft: randomized controlled trial (bmj.com)
“Evidence supporting the efficacy of parachutes is weak and guideline recommendations for their use are principally based on biological plausibility and expert opinion...To address these important gaps in evidence, we conducted the first randomized clinical trial of the efficacy of parachutes in reducing death and major injury when jumping from an aircraft.” A truly incredible satire.

Industry
The hottest advertising trend of 2018? Billboards. (thehustle.co)
“A decade ago, billboards were seen as outdated and irrelevant, relegated to informing you of the Cracker Barrel or gas station at the next exit. Today, in the midst of an online marketing boom, they’re one of the fastest-growing ad commodities on the market...If you’re stuck in traffic, the Google Play billboard looming over you on the freeway might suddenly advertise “soul soothing” music; likewise, a Target billboard might display rotisserie chickens, or other prepped meals, with a message like, ‘Running late? Keep dinner easy tonight.’”

Wealthfront: Silicon Valley Tech at Wall Street Prices (medium.com)
“Many people don’t have an intuitive grasp of the magic of compound interest, and so they certainly haven’t internalized the tyranny of compound fees. Then let us be clear: A 30-year old who invests $100,000 in his retirement with Wealthfront “for less than a night at the movies” will likely pay the company over $100,000 in fees by his 75th birthday.” Also interesting that Wealthfront just settled with regulators over misleading clients.

Following criticism, Robinhood backtracks on checking and savings feature (theverge.com)
“Earlier this week, stock-trading app Robinhood announced that it would offer... a checking and savings account that didn’t have any fees attached to it. But after the feature received intense criticism, the company appears to have backtracked, saying that it plans “to work closely with regulators as we prepare to launch our cash management program.””

Software
JIRA is an antipattern (techcrunch.com)
“Worst of all, though, is the endless implicit pressure for tickets to be marked finished, to be passed on to the next phase. Tickets, in the JIRA mindset, are taken on, focused on until complete, and then passed on, never to be seen again. They have a one-way life cycle: specification; design; development; testing; release. Doesn’t that sound a little … um … waterfall-y?”

Scaling engineering organizations (stripe.com)
“We’ve deliberately chosen to add engineers at a slower rate than the growth of of our user base and instead take an iterative approach to recruiting and developing our team. We were able to keep teams small, invest more in each person, and nimbly change our ways of working as needs changed. Since what works for fifty engineers may not work for ten or a thousand, we’ve built a culture of feedback and iterative thinking that enables us to continually make changes as we grow.”

Git Your SQL Together (with a Query Library) (caitlinhudon.com)
1) You will *always* need that query again, 2) Queries are living artifacts that change over time, 3) If it’s useful to you, it’s useful to others (and vice versa).

AI/Machine Learning
No time to read AI research? We summarized top 2018 papers for you (topbots.com)
So many papers...so little time.

Artificial Intelligence Index: 2018 Annual Report (aiindex.org)
“The AI Index is an effort to track, collate, distill, and visualize data relating to artificial intelligence. It aspires to be a comprehensive resource of data and analysis for policymakers, researchers, executives, journalists, and the general public to develop intuitions about the complex field of AI.”

Misc
Healthcare Triage: Do Standing Desks' Benefits Stand Up to Research? (youtube.com)
“Research suggests that warnings about sitting at work are overblown, and that standing desks are overrated as a way to improve health.”

Want to get in touch? Reply here, or visit nicholasbradford.io
Haven’t signed up yet? Click here.
Views expressed here are my own and not representative of any past or present affiliated entity.

2018-12-03 Tech Ingester: Taste, Dropping Dragonfly, and the Science of the Job Search

Tech Ingester

Best
Taste for Makers (paulgraham.com)
“Taste. You don't hear that word much now. And yet we still need the underlying concept, whatever we call it. What my friend meant was that he wanted students who were not just good technicians, but who could use their technical knowledge to design beautiful things.”

Industry
We are Google employees. Google must drop Dragonfly. (medium.com)
“Our opposition to Dragonfly is not about China: we object to technologies that aid the powerful in oppressing the vulnerable, wherever they may be.”

Facebook's story on Sheryl Sandberg and ongoing obsession with George Soros keeps evolving — here's the latest (cnbc.com)
“Definers pushed the idea that liberal financier Soros was behind a growing anti-Facebook movement in an effort to de-legitimize the campaign, the Times reported.”

Facebook is failing its black employees and its black users (facebook.com)
“The following memo was written and circulated by me to all of Facebook’s employees around the world. It was sent November 8, 2018, shortly before my final day at the company.”

The Science of the Job Search, Part VII: You Only Need 50% of Job “Requirements (talent.works)
“You’re as likely to get a job interview meeting 50% of job requirements as meeting 90% of them.”

Software
The faster you unlearn OOP, the better for you and your software (dpc.pw)
“...This oversimplified example with just 3 interacting classes is already becoming a typical OOP nightmare. A simple data transformation becomes a bunch of awkward, intertwined methods that call each other for no reason other than OOP dogma of encapsulation.”

5 Lessons Learned From Writing Over 300,000 Lines of Infrastructure Code (gruntwork.io)
“Building production-grade infrastructure is hard. And stressful. And time consuming. Very time consuming.”

cloud-nuke: how we reduced our AWS bill by ~85% (gruntwork.io)
“We set out to build a tool that periodically goes through our AWS account and deletes all idle resources.”

Why is 2 * (i * i) faster than 2 * i * i in Java? (stackoverflow.com)
“There is a slight difference in the ordering of the bytecode.”

Learn Blockchains by Building One (hackernoon.com)
“I like learning by doing. It forces me to deal with the subject matter at a code level, which gets it sticking. If you do the same, at the end of this guide you’ll have a functioning Blockchain with a solid grasp of how they work.”

Misc
Starting a Business in Silicon Valley (tlalexander.com)
“I lost $100,000 over five years on my first halfway successful venture. But as far as I'm concerned, it's been a success.”

Company Tried to Patent My Work After a Job Interview (patentpandas.org)
Hint: the company’s name rhymes with “Google”.

Kialo: A debate platform powered by reason (kialo.com)
“Kialo cuts through the noise typically associated with social and online media, making it easy to engage in focused discussion.”

Boss as a Service (bossasaservice.life)
“Hire a boss who will ensure you're consistently being productive.”

What I Learned From My Year-Long Minimalist Wardrobe Challenge (onlywhatmatters.com)
“The year long minimalist challenge started with the selection of 26 core items that I would wear for the entire year... And unlike the 100 day challenge, washing my clothing was allowed.”

Want to get in touch? Reply here, or visit nicholasbradford.io
Haven’t signed up yet? Click here.
Views expressed here are my own and not representative of any past or present affiliated entity.

11/26/2018 Tech Ingester: kids with smartphones, low asset returns, and the intellectual dark web

Tech Ingester

Best

The end of the beginning (ben-evans.com)
A presentation at a16z’s annual tech conference. “Close to three quarters of all the adults on earth now have a smartphone, and most of the rest will get one in the next few years. However, the use of this connectivity is still only just beginning. Ecommerce is still only a small fraction of retail spending, and many other areas that will be transformed by software and the internet in the next decade or two have barely been touched. Global retail is perhaps $25 trillion dollars, after all.”

What the Times got wrong about kids and phones (cjr.org)
“Some research does show that parents’ wealth and education are correlated with their limiting of kids’ screen time… However, parents with means are the ones who have the most resources to eschew media in favor of activities perceived as being of higher value (like by hiring nannies to occupy their children’s time…)... It’s a form of virtue signaling.”

Beating the Averages (paulgraham.com)
“I'll begin with a shockingly controversial statement: programming languages vary in power...This idea is rarely followed to its conclusion, though. After a certain age, programmers rarely switch languages voluntarily.”

Industry
Bridgewater's Dalio Sees Low Returns on Assets in Long Term (bloomberg.com)
“Today’s debt cycle is most reminiscent of the late 1930s, a time of rising populism and growing disparity between the rich and poor. Then and now, governments will have limited ability to deal with an economic downturn.”

Software Engineer Salary Data (triplebyte.com)
“Most online salary data is suspect because of self-reporting and selection bias. In contrast, we're able to aggregate all offers made to engineers we work with at Triplebyte. We work predominantly with Silicon Valley tech companies.”

Software
The C++ Build Process Explained (github.com)
“The C++ build process is built on top of the C build process which was hacked together in 1972 on a PDP-7 with at most 144 KB of RAM. Honestly, I'm surprised it aged so well.”

Amazon Web Services in Plain English (expeditedssl.com)
“Hey, have you heard of the new AWS services: ContainerCache, ElastiCast and QR72? Of course not, I just made those up. But with 50 plus opaquely named services, we decided that enough was enough and that some plain english descriptions were needed.”

Forecasting at Uber: An Introduction (uber.com)
“In recent years, machine learning, deep learning, and probabilistic programming have shown great promise in generating accurate forecasts. In addition to standard statistical algorithms, Uber builds forecasting solutions using these three techniques. Below, we discuss the critical components of forecasting we use, popular methodologies, backtesting, and prediction intervals.”

AI/Machine Learning
Introducing AdaNet: Fast and Flexible AutoML with Learning Guarantees (googleblog.com)
“We’re excited to share AdaNet, a lightweight TensorFlow-based framework for automatically learning high-quality models with minimal expert intervention. AdaNet builds on our recent reinforcement learning and evolutionary-based AutoML efforts to be fast and flexible while providing learning guarantees. Importantly, AdaNet provides a general framework for not only learning a neural network architecture, but also for learning to ensemble to obtain even better models.”

Acoustic Detection of Humpback Whales Using a Convolutional Neural Network (googleblog.com)
“We developed algorithms to identify humpback whale calls in 15 years of underwater recordings from a number of locations in the Pacific. The results of this research provide new and important information about humpback whale presence, seasonality, daily calling behavior, and population structure.”

Open Sourcing BERT: State-of-the-Art Pre-training for Natural Language Processing (googleblog.com)
“BERT builds upon recent work in pre-training contextual representations — including Semi-supervised Sequence Learning, Generative Pre-Training, ELMo, and ULMFit. However, unlike these previous models, BERT is the first deeply bidirectional, unsupervised language representation, pre-trained using only a plain text corpus.”

Misc
Color: From Hexcodes to Eyeballs (jamie-wong.com)
“Even if you’re a person who understands that most things are deeper than they look, color is way deeper than you would reasonably expect.”

Science Is Getting Less Bang for Its Buck (theatlantic.com)
“Despite vast increases in the time and money spent on research, progress is barely keeping pace with the past. What went wrong?”

The Voice of the ‘Intellectual Dark Web’ (politico.com)
“Claire Lehmann’s online magazine, Quillette, prides itself on publishing ‘dangerous’ ideas other outlets won’t touch. How far is it willing to go?”

Medium is a poor choice for blogging (medium.com)
“Sure, Medium editor is nice. Typography is good (for English, other languages are not supported). Publishing is free. So the ones who pay for all that are your readers. Basically, you’re selling readers to aggressive Medium self-advertising so that Medium could make laughably small money from a miserable fraction of them. And annoy the hell out of everyone else.”

If you want to understand Silicon Valley, watch Silicon Valley (gatesnotes.com)
“...like all great parodies it captures a lot of truths. Most of the different personality types you see in the show feel very familiar to me. The programmers are smart, super-competitive even with their friends, and a bit clueless when it comes to social cues. Personally, I identify most with Richard.”

Want to get in touch? Reply here, or visit nicholasbradford.io
Haven’t signed up yet? Click here.
Views expressed here are my own and not representative of any past or present affiliated entity.

11/12/2018 Tech Ingester: Unreadable Libraries, Another Red Hat, and Self-Driving Cars for Retirement Homes

Tech Ingester

Best
How one Florida retirement home became ground zero for self-driving cars (thehustle.co)
“The mega-development features 648 holes of golf, 100 pickleball courts, more than 100 restaurants, and 2700 social clubs… The Villages — the fastest growing metro area in the United States 4 of the last 5 years — has another advantage over West Palm Beach: it’s privately owned. Now, the family that runs it is sitting shotgun for the ride to a self-driving future.”

Industry
Why There Will Never Be Another Red Hat: The Economics of Open Source (a16z.com)
“Open source software powers the world’s technology...Yet, with all that momentum, there’s a vocal segment of software insiders that preach the looming failure of open source software against competition from proprietary software vendors.”

Google Plans Large New York City Expansion (wsj.com)
Interesting timing <cough cough Amazon cough cough>.

Software
Is the Scala 2.8 collections library a case of “the longest suicide note in history”? (stackoverflow.com)
“Those previously unfamiliar with Scala… now have to make sense of method signatures like: `def map[B, That](f: A => B)(implicit bf: CanBuildFrom[Repr, B, That]): That`.” Featuring an answer by Martin Odersky.

Five Questions About Language Design (paulgraham.com)
Questions/advice on what your guiding philosophy should be when designing a programming language. “A lot of people wish that hacking was mathematics, or at least something like a natural science. I think hacking is more like architecture. Architecture is related to physics, in the sense that architects have to design buildings that don't fall down, but the actual goal of architects is to make great buildings, not to make discoveries about statics.”

AI/Machine Learning
Curiosity and Procrastination in Reinforcement Learning (googleblog.com)
“Standard RL algorithms struggle with environments where feedback to the agent is sparse… we propose a novel episodic memory-based model of granting RL rewards, akin to curiosity, which leads to exploring the environment. Since we want the agent not only to explore the environment but also to solve the original task, we add a reward bonus provided by our model to the original sparse task reward.”

Horizon: Facebook’s Open Source Applied Reinforcement Learning Platform (fb.com)
“Horizon is an end-to-end platform designed to solve industry applied RL problems where datasets are large (millions to billions of observations), the feedback loop is slow (vs. a simulator), and experiments must be done with care because they don’t run in a simulator. Unlike other RL platforms, which are often designed for fast prototyping and experimentation, Horizon is designed with production use cases as top of mind.”

Misc
What I say to people who are looking for a job (gregkamradt.com)
Some very practical tips to common questions like “I don’t know what I want to do” or “my job is lame I think I want to do XYZ”

What if the Placebo Effect Isn’t a Trick? (nytimes.com)
“New research is zeroing in on a biochemical basis for the placebo effect — possibly opening a Pandora’s box for Western medicine.”

Want to get in touch? Reply here, or visit nicholasbradford.io
Haven’t signed up yet? Click here.
Views expressed here are my own and not representative of any past or present affiliated entity.

11/5/2018 Tech Ingester: Recruiting Madness, More Trolley Problems, and the Secret Restaurant Empire

Tech Ingester

Best
Should a self-driving car kill the baby or the grandma? Depends on where you’re from. (technologyreview.com)
“The infamous “trolley problem” was put to millions of people in a global study, revealing how much ethics diverge across cultures.”

A Dark Consensus About Screens and Kids Begins to Emerge in Silicon Valley (nytimes.com)
“Tim Cook, the C.E.O. of Apple, said earlier this year that he would not let his nephew join social networks. Bill Gates banned cellphones until his children were teenagers, and Melinda Gates wrote that she wished they had waited even longer. Steve Jobs would not let his young children near iPads.”

Industry
Goldman, JPMorgan Hit Pause on Intern Recruiting ‘Madness’ (wsj.com)
“Goldman Sachs Group Inc. and JPMorgan Chase & Co. won’t interview or extend summer internship offers to college sophomores this year and will go back to recruiting students in the fall of their junior year.”

Uber’s Secret Restaurant Empire (bloomberg.com)
“When the virtual restaurant team notices supply gaps in any given neighborhood—if, say, the data show that the number of brunch places is lower than could be served based on searches—they’ll begin contacting businesses in the area.”

Apple’s new iPad Pro has Face ID, USB-C, and slimmer bezels than ever before (theverge.com)
“Preorders start today, and it goes on sale November 7th.”

Software
Microservices Are Something You Grow Into, Not Begin With (nickjanetakis.com)
“As we skim HackerNews and other programming news outlets we often come across tech posts from Google, Netflix, Amazon and Facebook and they love to talk about how many hundreds or thousands of services they run and go over the benefits of doing things their way… But let’s face it. You probably don’t have 1,000 developers working on a massive project with 10+ years of history.”

Do We Worship Complexity? (innoq.com)
“A particularly blatant case of complexity worshiping is the statement “This doesn’t work for us. Our challenges are much greater than those of Amazon or Google.” I’ve heard that from employees of different companies. Such statements are surprising: companies like Amazon or Google have extremely complex IT systems. Their economic success depends directly on these IT systems. Not least because of these IT systems, they are among the most valuable companies in the world.”

Lazy Vals in Scala: A Look Under the Hood (codecentric.de)
“Scala allows the special keyword lazy in front of val in order to change the val to one that is lazily initialized. While lazy initialization seems tempting at first, the concrete implementation of lazy vals in scalac has some subtle issues.”

Why Robinhood uses Airflow (robinhood.engineering)
“Airflow, developed at Airbnb has a growing community and seemed to be the best suited for our purposes. It is a horizontally scalable, distributed workflow management system which allows us to specify complex workflows using Python code.”

AI/Machine Learning
Understanding the Bias-Variance Tradeoff (fortmann-roe.com)
One of the better explanations I’ve found so far.

Misc
Talking to the mailman (Interview by Rob Lucas) (newleftreview.org)
A really interesting long-form interview with Richard Stallman.

iPhones are hard to use (fawny.org)
For lots of reasons: “My almost-blind friend upgraded from an iTouch to an iPhone 8, then couldn’t check his voicemail for weeks because iPhone keyboards and keypads randomly change or invert their colours and he simply could not see or locate the number buttons.”

Comparison of Engineering Levels and Compensation (levels.fyi)
Compare and contrast the different levels and comp across top tech companies.

Justice Department Charges Ex-Goldman Bankers in Malaysia 1MDB Scandal (wsj.com)
“Two senior Goldman Sachs bankers paid bribes and stole and laundered money from a Malaysian sovereign-wealth fund, U.S. prosecutors allege, putting the bank at the center of one of the biggest financial frauds in history.”

Want to get in touch? Reply here, or visit nicholasbradford.io
Haven’t signed up yet? Click here.
Views expressed here are my own and not representative of any past or present affiliated entity.

10/22/2018 Tech Ingester: Software Consulting, Robinhood Clearing, and Generative Algorithms

Tech Ingester

Best
The Key to Becoming a Software Consultant (daedtech.com)
Counter-intuitively, “Don’t ever let would-be consulting clients pay you for code that you write.” Less about consulting, and more about the best way to position yourself to provide thought leadership instead of merely skilled labor.

Programming Bottom-Up (paulgraham.com)
“The traditional approach is called top-down design… Experienced Lisp programmers divide up their programs differently. As well as top-down design, they follow a principle which could be called bottom-up design-- changing the language to suit the problem...Instead of a single, monolithic program, you will get a larger language with more abstract operators, and a smaller program written in it. “

Industry

Software
Under the Hood of Clearing by Robinhood (robinhood.engineering)
“Over 40% of banking systems still operate on COBOL, a language developed over 60 years ago when computers took up an entire room — and 92 of the top 100 banks are still using mainframe computers. Broadly, the brokerage and clearing industries are no different. Until now.”

Capturing and enhancing in situ system observability for failure detection (acolyer.org)
“Today the state of the practice is to log and try to recover from a failed call at the client, while a totally separate failure detection infrastructure is responsible for figuring out whether or not things are working as desired. What Panorama does is turn clients into observers and reporters of the components they call, using these observations to determine component health.”

AI/Machine Learning
Same-different problems strain convolutional neural networks (acolyer.org)
“The CNNs really struggled on problems where the abstract rule required detecting whether things were the same or different (congruent up to some transformation), whereas they achieved good accuracy on spatial relation problems.” The authors hypothesize that “learning templates for arrangements of objects becomes rapidly intractable because of the combinatorial explosion in the number of features to be stored.”

Learning Math for Machine Learning (ycombinator.com)
“It turns out that a lot of people — including engineers — are scared of math. To begin, I want to address the myth of “being good at math.” The truth is, people who are good at math have lots of practice doing math. As a result, they’re comfortable being stuck while doing math.”

3 facts about time series forecasting that surprise experienced machine learning practitioners. (towardsdatascience.com)
“Time series forecasting… is one of the most applied data science techniques in business, used extensively in finance, in supply chain management and in production and inventory planning, and it has a well established theoretical grounding in statistics and dynamic systems theory. Yet it retains something of an outsider status compared to more recent and popular machine learning topics such as image recognition and natural language processing..”

Misc
67% of workers earning over $100,000 see themselves quitting in the next six months—here's why (cnbc.com)
“According to Kropp, the average increase in compensation for a worker who quits their old job for a new one in today's tight labor market is about 15 percent”

On Generative Algorithms (inconvergent.net)
“I've been experimenting quite a lot with recreating various biological behaviours. Part of the challenge is to try to recreate some pattern or behaviour with as few and as simple rules as possible.” Some of them turn out really funky and beautiful.

Want to get in touch? Reply here, or visit nicholasbradford.io
Haven’t signed up yet? Click here.
Views expressed here are my own and not representative of any past or present affiliated entity.

10/15/2018 Tech Ingester: The Big Hack, Paige.AI uproar, and Bragging

Tech Ingester

Best
The Big Hack: How China Used a Tiny Chip to Infiltrate U.S. Companies (bloomberg.com)
Bloomberg claims “The attack by Chinese spies reached almost 30 U.S. companies, including Amazon and Apple, by compromising America’s technology supply chain, according to extensive interviews with government and corporate sources.” This has been met by strong denials from the companies involved - read more about the story’s evolution.

Industry
a16z Podcast: What’s at the Core of the Latest Apple Announcement? (a16z.com)
“With something that’s gone from toy to phone to fashion item — and just pivoted to a health monitor that can literally save lives — where are we now? How closely aligned is health to the overall value proposition, and what are some of the characteristics of how Apple innovates as a company as a whole… from components and building blocks to how it all comes together?”

Why I Quit Google to Work for Myself (mtlynch.io)
“My first denied promotion taught me the wrong lesson. I thought I could keep doing the same work but package it to look good for the promotion committee. I should have done the opposite: figure out what the promotion committee wants, and do that work exclusively.”

Sloan Kettering’s Cozy Deal With Start-Up Ignites a New Uproar (nytimes.com)
“Hospital pathologists have strongly objected to the Paige.AI deal, saying it is unfair that the founders received equity stakes in a company that relies on the pathologists’ expertise and work amassed over 60 years.”

The Facebook Security Meltdown Exposes Way More Sites than Facebook (wired.com)
“Facebook revealed that it had suffered a security breach that impacted at least 50 million of its users, and possibly as many as 90 million. What it failed to mention initially, but revealed in a followup call Friday afternoon, is that the flaw affects more than just Facebook. If your account was impacted it means that a hacker could have accessed any account that you log into using Facebook.”

The demise of Rethink Robotics shows how hard it is to make machines truly smart (technologyreview.com)
“It might seem like a bad sign for robotics and AI that Rethink Robotics, which pioneered the development of more intelligent and user-friendly robots, plans to shut down. But the true picture is more complicated.”

Software
UML Diagram Types Guide: Learn About All Types of UML Diagrams with Examples (creately.com)
All of them? Yes, all of them.

AI/Machine Learning
Detecting spacecraft anomalies using LSTMs and nonparametric dynamic thresholding (acolyer.org)
“The major issue being worked on now is to reduce the number of false positives (the bane of many an alerting system!): “Investigation of even a couple of false positives can deter users and therefore achieving high precision with over a million telemetry values being processed per day is essential for adoption.”

Misc
The Best Article Ever Written About Bragging (lesspenguiny.com)
“I wanted to eject from the conversation. In fact, I never wanted to see this guy’s face again — not even in a weird, clinical “how much bragging can I endure” way. So I left, genuinely curious about whether this guy’s own mother returned his calls.”

From Private To Public: How To Read An S-1 (crunchbase.com)
“Maybe coverage of a hot IPO contains excellent information on how much money the CEO stands to make when her company debuts, but zero details on its customer cohorts... If you could do your own research, you’d be in business.”

A diet guru explains why you should eat dinner at 2pm (qz.com)
“There have been two main changes in dietary habits from the 1970s (before the obesity epidemic) until today. First, there was the change is what we were recommended to eat. Prior to 1970, there was no official government sanctioned dietary advice… But the other major change was in when we eat.”

Want to get in touch? Reply here, or visit nicholasbradford.io
Haven’t signed up yet? Click here.
Views expressed here are my own and not representative of any past or present affiliated entity.

10/1/2018 Tech Ingester: Bezos Unbound, Why I’m Done with Chrome, and the SEC

Tech Ingester

By the way, Elon Musk settled the SEC charges - Musk will step down as Chairman, and he and Tesla pay a $40mm penalty.

Best
Bezos Unbound: Exclusive Interview With The Amazon Founder On What He Plans To Conquer Next (Forbes)
‘"There are different businesses where the market is limited," adds the man whose company should hit $210 billion in revenue this year. "But we just don't have that issue."’

Industry
Tesla, software and disruption (Benedict Evans’ blog)
An analysis of what does and doesn’t make Tesla disruptive - discusses batteries, software, experience, and autonomy.

Dozens at Facebook Unite to Challenge Its ‘Intolerant’ Liberal Culture (NYT)
“We are a political monoculture that’s intolerant of different views,” Brian Amerige, a senior Facebook engineer, wrote in the post... “We claim to welcome all perspectives, but are quick to attack — often in mobs — anyone who presents a view that appears to be in opposition to left-leaning ideology.”

Software
Why I’m Done With Chrome (Cryptography Engineering blog)
“A few weeks ago Google shipped an update to Chrome that fundamentally changes the sign-in experience. From now on, every time you log into a Google property (for example, Gmail), Chrome will automatically sign the browser into your Google account for you. It’ll do this without asking, or even explicitly notifying you.”

How One Website Exploited Amazon S3 to Outrank Everyone on Google (blog.usejournal.com)
A dive into the World of SEO, Affiliate Marketing, Amazon S3, and why so many thousands of sites claim to want to give you free coupons that don’t work.

Elements of Programming Style - Brian Kernighan (YouTube)
Kernighan reviews a variety of examples, many from programming textbooks claiming to be good, to show principles such as: don’t be too clever, know your language and its idioms, make every comment count, do something after every decision, and more.

An Intensive Introduction to Cryptography (intensecrypto.org)
From Harvard professor Boaz Barak: “These are lecture notes for lecture notes for an introductory but fast-paced undergraduate/beginning graduate course on cryptography.”

AI/Machine Learning
Learn Deep Reinforcement Learning in Depth in 60 days (GitHub)
“This repository wants to guide you through the Deep Reinforcement Learning algorithms, from the most basic ones to the highly advanced AlphaGo Zero. You will find the main topics organized by week and the resources suggested to learn them.”

Dopamine (GitHub)
“Dopamine is a research framework for fast prototyping of reinforcement learning algorithms. It aims to fill the need for a small, easily grokked codebase in which users can freely experiment with wild ideas (speculative research).” By Google.

Misc
Computer Or Human? + Thad (NPR)
“Alix and Lulu begin with the story of a man who challenges the machine that gave him a speeding ticket. Then we hear from Thad Starner, who has been wearing a computer since 1993.”

Why Socialists Don't Believe In Fun (orwell.ru)
A 1943 essay from George Orwell about the difficulty writers face when describing ideas of happiness, perfection, and utopia.

Want to get in touch? Reply here, or visit nicholasbradford.io
Haven’t signed up yet? Click here.
Views expressed here are my own and not representative of any past or present affiliated entity.

9/10/2018 Tech Ingester: Augmenting Long-Term Memory, AGI Literature Review, and Musk’s Podcast Appearance

Tech Ingester

Best
Augmenting Long-Term Memory (Augmenting Cognition)
Michael Nielsen (fellow at Y Combinator Research) reviews general personal-memory system and details his use of the Anki system, an automated flashcard system with exponential backoff to encourage building long-lasting memories of key facts.

Productivity (Google Docs)
A collection of miscellaneous productivity tips and tricks, distilled down to 5 pages (spoiler alert: I wrote it to summarize the material I give in my office lunch ‘n learns).

Industry
Tesla stock closes down 6% after top executives resign and Elon Musk smokes weed on video (CNBC)
“How does that work? Do people get upset at you if you do certain things?” asked Joe Rogan on the podcast while Musk smoked. Yes. Yes they do. They also make a lot of memes as a result.

Software
Morning Cup of Coding (Human Readable)
A daily programming newsletter featuring long form technical articles of all fields of software engineering.

React-howto (GitHub)
From one of the original React creators: advice for navigating the [confusing] ecosystem, with an order of prioritization.

AI/Machine Learning
Deep Code Search (The Morning Paper)
“DeepCS is just such a search engine for code... During training, it takes code snippets (methods) and corresponding natural language descriptions (from the method comments) and learns a joint-embedding. I.e., it learns embeddings such that a method description and its corresponding code snippet are both mapped to a similar point in the same shared embedding space. Then given a natural language query, it can embed the query in vector space and look for nearby code snippets.”

AGI [Artificial General Intelligence] Safety Literature Review (arXiv)
“The intention of this paper is to provide an easily accessible and up-to-date collection of references for the emerging field of AGI safety. A significant number of safety problems for AGI have been identified. We list these, and survey recent research on solving them. We also cover works on how best to think of AGI from the limited knowledge we have today, predictions for when AGI will first be created, and what will happen after its creation. Finally, we review the current public policy on AGI.”

Misc
What's the Deal with Steam Rising from the NYC Streets? (CityLab)
“The steam pouring out is the very kind used to clean the dishes in a New York restaurant, sterilize hospital equipment, and heat up cheese curds in artisan shops. Con Edison operates the world’s largest network of steam pipes… It’s bigger than the next nine largest steam systems combined.”

Want to get in touch? Reply here, or visit nicholasbradford.io
Haven’t signed up yet? Click here.
Views expressed here are my own and not representative of any past or present affiliated entity.

8/27/2018 Tech Ingester: Bad Keynote Speakers, Interpretable ML, and Dijkstra’s in Disguise

Tech Ingester

Note: no newsletter next week (I’m off on vacation), but in the meantime I recommend you amuse yourself with The Morning Brew for daily industry/finance/tech news.

Best
Q: Why Do Keynote Speakers Keep Suggesting That Improving Security Is Possible? (YouTube)
“I will explain why the proliferation of ubiquitous technology is good in the same sense that ubiquitous Venus weather would be good, i.e., not good at all. Using case studies involving machine learning and other hastily-executed figments of Silicon Valley’s imagination, I will explain why computer security… are difficult to achieve if developers insist on literally not questioning anything that they do.”

The Great AI Debate - "Is Interpetability necessary for machine learning?" NIPS2017 (YouTube)
For context on the debate, check out “The Mythos of Model Interpretability” (arXiv) and “Towards A Rigorous Science of Interpretable Machine Learning” (arXiv). Arguments for interpretability include the fact that many domains require high standards (e.g. health, finance), that without a casual model it’s extremely difficult to avoid learning spurious correlations due to biased training/test sets, and without interpretability it’s impossible to determine when the algorithm diverges from human intuition or violates common sense. Arguments against include that most ML domains rarely require a detailed explanation, an accurate model trumps an explainable model every time, most ML models are actually somewhat interpretable already, and humans have a long history of using methods that work empirically (i.e. health, engineering) before understanding underlying theory.

Industry
The Future of Notebooks: Lessons from JupyterCon (Will Crichton’s blog)
1) “Reactive” notebooks which track dependencies between cells to make re-running easier, 2) Jupyter is the new Bash and is increasingly being used for rich and easily diagnosable automation, and 3) Data science is a “gateway drug” in education.

What Do You Believe Now That You Didn't Five Years Ago? Centralized Wins. Decentralized Loses. (High Scalability)
“You know it's bad when GitHub managed to recentralize an inherently distributed system like git."

Software
Dijkstra's in Disguise (Eric Jang’s blog)
“It turns out that many algorithms I've encountered in my computer graphics, finance, and reinforcement learning studies are all variations of this relaxation principle in disguise. It's quite remarkable (embarrassing?) that so much of my time has been spent on such a humble technique taught in introductory computer science courses!”

AI/Machine Learning
Now anyone can train Imagenet in 18 minutes (Fast.ai)
Significant as proof that even researchers on a budget can now train large models at low cost. Their team “managed to train Imagenet to 93% accuracy in just 18 minutes, using 16 public AWS cloud instances, each with 8 NVIDIA V100 GPUs, running the fastai and PyTorch libraries. This is a new speed record... and costs around $40 to run.”

D. E. Shaw Group Forms New Machine Learning Research Group (PR Newswire)
“The new Machine Learning Research Group, which will operate in parallel to the firm's longstanding machine learning efforts, will be overseen by Dr. Pedro Domingos.”

TensorFlow 2.0 is Coming (Google Groups)
Highlights we can expect: eager execution, support for more platforms and languages, and cleaning of the duplicated/deprecated APIs.

Misc
A Bunch of Men Got Tinder-Pranked in Union Square (NYT)
“Aponte used the dating app to invite men to meet her in New York City's Union Square last Saturday. When they all got there at the same time, Aponte revealed that they were actually invited to take part in her public dating competition.” - (CBS)

Stacking concrete blocks is a surprisingly efficient way to store energy (QZ)
“A startup called Energy Vault thinks it has a viable alternative to pumped-hydro: Instead of using water and dams, the startup uses concrete blocks and cranes.”

Want to get in touch? Reply here, or visit nicholasbradford.io
Haven’t signed up yet? Click here.
Views expressed here are my own and not representative of any past or present affiliated entity.

8/20/2018 Tech Ingester: Magic Leap, the Power of Less, and Google’s Tracking

Tech Ingester

Best
The 10:1 rule of writing and programming (Yevgeniy Brikman’s blog)
“The ratio of “raw materials” to “finished product” in a book is roughly 10:1. Keep this in mind the next time an editor asks you for a timeline! If you want to write a 300 page book, you’ll probably have to write around 3,000 pages.”

After almost a decade and billions in outside investment, Magic Leap's first product is finally on sale (CNBC)
“Think of all the software you typically use — like email and a video player — floating in the real world in front of you. Since you can see really far through the headset, that software can pretty much exist anywhere within your eyesight. You could have a digital chat with a 3-D avatar of another user right in your living room, or go sit in a football stadium and drop a life-sized T. rex out on the field.”

Industry
Google tracks your movements, like it or not (AP)
“An Associated Press investigation found that many Google services on Android devices and iPhones store your location data even if you’ve used privacy settings that say they will prevent it from doing so.”

Elon Musk Details ‘Excruciating’ Personal Toll of Tesla Turmoil (NYT)
“In an hourlong interview with The New York Times, he choked up multiple times, noting that he nearly missed his brother’s wedding this summer and spent his birthday holed up in Tesla’s offices as the company raced to meet elusive production targets on a crucial new model.”

Software
The System Design Primer (GitHub)
Everything you could want to know about system design (web in particular), with lots of resources for interview prep.

Docker for Beginners (Docker-Curriculum)
A very well put-together introduction to Docker that’ll quickly have you up and running your own multi-container web app using AWS.

AI/Machine Learning
When Recurrent Models Don't Need to be Recurrent (Off The Convex Path)
“Feed-forward models can offer improvements in training stability and speed, while recurrent models are strictly more expressive. Intriguingly, this added expressivity does not seem to boost the performance of recurrent models. Several groups have shown feed-forward networks can match the results of the best recurrent models on benchmark sequence tasks.”

Natural Language Processing is Fun! (Medium)
“Doing anything complicated in machine learning usually means building a pipeline. The idea is to break up your problem into very small pieces and then use machine learning to solve each smaller piece separately.”

Misc

The Power of Less: Changing Behavior with Leo Babauta (Tim Ferriss’ blog)
“The only way you’ll form long-lasting habits is by applying the Power of Less: focus on one habit at a time, one month at a time, so that you’ll be able to focus all your energy on creating that one habit.”

Librarian (Fermat’s Library)
“A Chrome extension that enhances arXiv papers. Get direct links to references, BibTeX extraction and comments on all arXiv papers”

Want to get in touch? Reply here, or visit nicholasbradford.io
Haven’t signed up yet? Click here.
Views expressed here are my own and not representative of any past or present affiliated entity.

8/6/2018 Tech Ingester: Amazon dark patterns, Production Minimalism, and Cutting through the AutoML hype

Tech Ingester

Best
What do machine learning practitioners actually do? (Fast.ai)
Part 1 of a three-part series: “For myself and many others I know, I would highlight two of the most time-consuming and frustrating aspects of machine learning (in particular, deep learning) as: 1) Dealing with data formatting, inconsistencies, and errors is often a messy and tedious process. 2) Training deep learning models is a notoriously brittle process right now.” Check out Part 2 on An Opinionated Introduction to AutoML and Neural Architecture Search, and Part 3 Google's AutoML: Cutting Through the Hype.

Industry
Amazon dark patterns ('Net Instructions)
“I wouldn't be this annoyed if this wasn't the second time in the last year I tried leaving a less then perfect review on a product I had purchased on Amazon, only to hear that my review "could not be posted to the website in its current form" after I spent half an hour of my time reviewing (with pictures) it.”

Uber’s self-driving trucks division is dead, long live Uber self-driving cars (TechCrunch)
“Rather than having two groups working side by side, focused on different vehicle platforms, I want us instead collaborating as one team,” according to an email reviewed by TechCrunch that was sent by [Uber Advanced Technologies Group head] Meyhofer to employees. “...We believe delivering on self-driving for passenger applications first, and then bringing it to freight applications down the line, is the best path forward. For now, we need the focus of one team, with one clear objective.”

Software
In Pursuit of Production Minimalism (Brandur)
“Antoine de Saint Exupéry, a French poet and pioneering aviator, had this to say on the subject: ‘It seems that perfection is reached not when there is nothing left to add, but when there is nothing left to take away.’ Most of us can benefit from architecture that’s a little simpler, a little more conservative, and a little more directed.”

AI/Machine Learning
Google AI Chief Jeff Dean’s ML System Architecture Blueprint (Medium)
“Dean identifies six issues that impact ML hardware design within this five-year window, from purely architectural to mostly ML-driven concerns, including: Training, Batch Size, Sparsity and Embedding, Quantization and Distillation, Networks with Soft Memory, and Learning to Learn.”

Amazon’s Face Recognition Falsely Matched 28 Members of Congress With Mugshots (ACLU)
This generated a lot of buzz after the ACLU used Amazon Rekognition to compare Congress members with a mugshot database. Read Amazon’s response in “Thoughts On Machine Learning Accuracy” (AWS News Blog), which points out that the ACLU didn’t follow “best practices”.

Misc
Why Americans Spend So Much on Health Care—In 12 Charts (WSJ)
“Americans aren’t buying more health care overall than other countries. But what they are buying is increasingly expensive. Among the reasons is the troubling fact that few people in health care, from consumers to doctors to hospitals to insurers, know the true cost of what they are buying and selling.”

Here's How America Uses Its Land (Bloomberg)
Turns out 41% is devoted to livestock. And, “On a percentage basis, urban creep outpaces growth in all other land-use categories. Another growth area: land owned by wealthy families. According to The Land Report magazine, since 2008 the amount of land owned by the 100 largest private landowners has grown from 28 million acres to 40 million, an area larger than the state of Florida.”

Wall Street Addresses #MeToo Concerns With 'the Weinstein Clause' (Fortune)
“Advisers are adding language to mergers that require one to legally vouch for the positive behavior of a company’s leadership. In some cases, merger agreements say buyers have the right to receive refunds if, after the deal, revelations of misbehavior damage business.”

Want to get in touch? Reply here, or visit nicholasbradford.io
Haven’t signed up yet? Click here.
Views expressed here are my own and not representative of any past or present affiliated entity.

7/30/2018 Tech Ingester: The Crackdown on Free Food, a Journey Into The Borg, and a 42,000-year-old Creature

Tech Ingester

Best
Why Google Stores Billions of Lines of Code in a Single Repository (ACM)
A 2016 article describing implementation details, pros, and cons of the Monorepo. “At Google, we have found, with some investment, the monolithic model of source management can scale successfully to a codebase with more than one billion files, 35 million commits, and thousands of users around the globe.”

The Best Textbooks on Every Subject (LessWrong)
“For years, my self-education was stupid and wasteful. I learned by consuming blog posts, Wikipedia articles, classic texts, podcast episodes...I've since discovered that textbooks are usually the quickest and best way to learn new material.”

Industry
San Francisco Bay Area cities are cracking down on free food at Facebook and other tech companies (Business Insider)
“Mountain View… will not allow a new office development where Facebook is set to move this fall to have a cafeteria with free food for employees. The restriction aims to increase business for local food retailers. San Francisco… is proposing a similar rule that would ban new workplace cafeterias for the same reason.”

Ex-Valve employee describes ruthless internal politics at 'self-organizing' companies (PC Gamer)
“According to Geldreich, the company in question leaked its friendly employee handbook as a PR move, but was actually a stressful, difficult place to work, full of backstabbing and manipulation in pursuit of bonuses and job security.”

Software
Into the Borg – SSRF inside Google production network (OpnSec)
“Google VRP rewarded me with $13,337, which corresponds to something like unrestricted file access!”

First Steps with Python Type System (Daftcode Blog)
Using the new type annotations + one of various type checkers, you can get your Python code to look and feel just like Scala code… looks like people aren’t such a fan of duck typing after all.

AI/Machine Learning
Machine Learning Guides (Google Developers)
“Simple step-by-step walkthroughs to solve common machine learning problems using best practices.”

Seedbank — discover machine learning examples (Medium)
Google’s “Seedbank now provides a place to search for Colab-powered Machine Learning examples. You can use the top-level categories to narrow your exploration and search for keywords inside of notebooks. Each seed has a preview that lets you quickly assess if you want to explore further. Once you click through to the Colab notebook, you’ll be immediately connected to a GPU kernel and can start working through the example or tutorial.”

Google is baking machine learning into its BigQuery data warehouse (TechCrunch)
“Using BigQuery ML, developers can build models using linear and logistical regression right inside their data warehouse without having to transfer data back and forth as they build and fine-tune their models. And all they have to do to build these models and get predictions is to write a bit of SQL.”

Misc
From the Poker Table to Wall Street (NYT)
“By the time she announced her retirement from pro poker… Ms. Selbst had won three World Series bracelets, netted $11.9 million in tournament winnings and spent a brief spell in 2015 as No. 1 in the Global Poker Index ranking… As her Twitter profile now slyly alludes — ‘I used to gamble and wake up late; now I gamble and wake up early’”.

Worms frozen in permafrost for up to 42,000 years come back to life (Siberian Times)
“Some 300 prehistoric worms were analysed - and two ‘were shown to contain viable nematodes’... Currently the nematodes are the oldest living animals on the planet.”

YC’s 2018 Summer Reading List (YCombinator blog)
Top picks include The Years of Lyndon Johnson, The Three Body Problem, Taming the Sun: Innovations to Harness Solar Energy and Power the Planet, Clean Meat: How Growing Meat Without Animals Will Revolutionize Dinner and the World, and more.

Want to get in touch? Reply here, or visit nicholasbradford.io
Haven’t signed up yet? Click here.
Views expressed here are my own and not representative of any past or present affiliated entity.

7/23/2018 Tech Ingester: The Wrong Abstraction, Knowledge Work in the Cloud, and Goodbye Microservices

Tech Ingester

Best
The Wrong Abstraction (Sandi Metz’s blog)
“If you find yourself passing parameters and adding conditional paths through shared code, the abstraction is incorrect. It may have been right to begin with, but that day has passed... the best strategy is to re-introduce duplication and let it show you what's right.”

Industry
Fin: How Knowledge Work is Becoming a Cloud Service (Google Docs)
A brief pitch presentation from the Fin personal assistant service on the benefits of an on-demand “executive assistant in the cloud” vs hiring one full-time, or trying to do it yourself (e.g. scheduling meetings, booking plane tickets, etc.).

What Industry Has The Highest Revenue Per Employee? (Craft)
An analysis of the RPE of S&P 500 companies and sectors. Spoiler alert: it’s energy. The winner, Valero, makes an incredible $9.4mm/employee; meanwhile Netflix makes $2.1mm, and Facebook $1.6mm. Friendly reminder that this is for revenue, not profit.

Software
Goodbye Microservices: From 100s of problem children to 1 superstar (Segment blog)
“Instead of enabling us to move faster, the small team found themselves mired in exploding complexity… Eventually, the team found themselves unable to make headway, with 3 full-time engineers spending most of their time just keeping the system alive..” They ended up moving back to a Monorepo.

AI/Machine Learning
The Building Blocks of Interpretability (Distill.pub)
“Interpretability techniques are normally studied in isolation. [In this article, Google Brain researchers] explore the powerful interfaces that arise when you combine them — and the rich structure of this combinatorial space.” As you would hope, comes with a variety of well-crafted visualizations.

DeepTest: automated testing of deep-neural-network-driven autonomous cars (The Morning Paper)
“Just like traditional software, DNN-based software...often demonstrate incorrect/unexpected corner-case behaviours that lead to dangerous consequences like a fatal collision...When used to test three of top performing DNNs from the Udacity self-driving car challenge, it unearthed thousands of erroneous behaviours.”

Misc
Flatiron District Fears Contamination After Steam Pipe Explosion (NYT)
“A day after a steam pipe exploded beneath Fifth Avenue in Manhattan, showering the Flatiron district with asbestos-filled muck, blocks of the neighborhood remained cordoned off Friday.”

Creativity, Inc. key takeaways (Medium)
Summary of the book about “Overcoming the Unseen Forces That Stand in the Way of True Inspiration” by Pixar co-founder and current President Ed Camull.

Want to get in touch? Reply here, or visit nicholasbradford.io
Haven’t signed up yet? Click here.
Views expressed here are my own and not representative of any past or present affiliated entity.

7/16/2018 Tech Ingester: DIY Web Crawlers, Neuroevolution, and the Great CEO Within

Tech Ingester

Best
The Great CEO Within (Google Docs)
A great 100-page free book by Matt Mochary on everything you need to be a first-time founder/CEO: individual habits and productivity, organizational structure, infrastructure, and miscellaneous processes to install.

Industry

The open-plan office is a terrible, horrible, no good, very bad idea (Medium)
“We’re squandering human health and potential on an epic scale by forcing the vast majority of people who dislike or hate the open office into that configuration. Their work deteriorates, their job satisfaction declines.”

Software
How to crawl a quarter billion webpages in 40 hours (Michael Nielsen’s blog)
“More precisely, I crawled 250,113,669 pages for just under 580 dollars in 39 hours and 25 minutes, using 20 Amazon EC2 machine instances. I carried out this project because (among several other reasons) I wanted to understand what resources are required to crawl a small but non-trivial fraction of the web. In this post I describe some details of what I did. Of course, there’s nothing especially new: I wrote a vanilla (distributed) crawler, mostly to teach myself something about crawling and distributed computing.”

How SQL Database Engines Work (YouTube)
A 30-minute lecture (plus Q&A) from the creator of SQLite on the internal structure of database engines.

AI/Machine Learning
Evolution Strategies as a Scalable Alternative to Reinforcement Learning (OpenAI blog)
“We’ve discovered that evolution strategies (ES)... rivals the performance of standard reinforcement learning (RL) techniques on modern RL benchmarks (e.g. Atari/MuJoCo)... ES is simpler to implement (there is no need for backpropagation), it is easier to scale in a distributed setting, it does not suffer in settings with sparse rewards, and has fewer hyperparameters. This outcome is surprising because ES resembles simple hill-climbing in a high-dimensional space based only on finite differences along a few random directions at each step.”

Misc
Wall Street Is Moving, and It’s Reshaping New York (Bloomberg)
Everyone’s packing up their Midtown and Wall Street offices and moving to the less costly Hudson Yards.

Trader builds $5 billion position after realizing he wasn't using test program, but brokerage says he can't keep his profit (CNBC)
“Harouna Traore, a day trader in France, thought he was using a demonstration version of British brokerage firm Valbury Capital's platform... [he] racked up a loss of more than $1 million before he realized the trades were real and, instead of quitting, he kept trading for a profit of 10 million euros, or $11.6 million.”

Casper opens a storefront for $25 naps (TechCrunch)
The Dreamery in NoHo will let you take a 45-minute nap on a Casper mattress.

Want to get in touch? Reply here, or visit nicholasbradford.io
Haven’t signed up yet? Click here.
Views expressed here are my own and not representative of any past or present affiliated entity.

7/9/2018 Tech Ingestor: Plumbing, Bayesian Recruiting, and the Puzzle Around You

Tech Ingestor

Best
The Bulk of Software Engineering in 2018 is Just Plumbing (Karl L. Hughes’ blog)
“This happens all the time during job-screening. We put engineers through rigorous screening processes and ask them intellectually stimulating questions, only to hire them and put them into the admittedly dull task of wiring up 5 or 6 services and making the screen look pretty.”

Industry
Grow the Puzzle Around You (Jessica Livingston’s blog)
“I wanted to tell you my own story. If you only know about me through the media, you might get the impression that my contribution to Y Combinator is that I’m Paul Graham’s wife...there's a bit more to the story than that.”

The impact of the ‘open’ workspace on human collaboration (The Royal Society)
“Rather than prompting increasingly vibrant face-to-face collaboration, open architecture appeared to trigger a natural human response to socially withdraw from officemates and interact instead over email and IM [a decrease of about ~70%].”

Bayesian Inference for Hiring Engineers (TripleByte)
“This is a messy world of overlapping signals and biases. And the recruiters and hiring managers making these decisions are guided primarily by pattern matching. We can do better... Rather than thinking of each technical screening step as crisp yes / no test, we should think of them as evidence updating our prior over a candidate’s skill.”

Software
Understanding F[_] in Scala (Medium)
Consider that `5` is a value of proper type `Int`. Typical generics such as `List` are first order types with type constructor `List[ _ ]` that takes a type like `Int` and returns a proper type `List[Int]`. To abstract over type constructors we have higher-kinded types; e.g. a `Mappable[F[ _ ]]` where `F[ _ ]` is some container such as `List[Int]` which can be `mapped` over. Also check out this Stack Overflow answer for a well-structured short example.

AI/Machine Learning
Ways to think about machine learning (Benedict Evans’ blog)
“...This gets to the heart of the most common misconception that comes up in talking about machine learning - that it is in some way a single, general purpose thing, on a path to HAL 9000, and that Google or Microsoft have each built *one*, or that Google 'has all the data', or that IBM has an actual thing called ‘Watson’.”

Misc
How a Hacker Proved Cops Used a Secret Government Phone Tracker to Find Him (Politico)
“The Hacker was found out through the warrantless use of a secretive surveillance technology known as a stingray, which snoops on cell phones… their use has expanded with little oversight—and no public knowledge that they were even being used.”

Want to get in touch? Reply here, or visit nicholasbradford.io
Haven’t signed up yet? Click here.
Views expressed here are my own and not representative of any past or present affiliated entity.

7/2/2018 Tech Ingestor: Control Groups, Coding Black Mirror, and Sleepwalking Through Life

Tech Ingestor

Best
Re-coding Black Mirror, Parts I-V (The Morning Paper)
If the popular sci-fi-gone-wrong show Black Mirror has ever made you anxious about the direction of technology, this series of paper summaries will show you just how close we are to making those nightmares a reality. Parts I (re-creating a virtual “you” after your death), II (automatic emotion recognition, social credit systems), III (misuse of fitness wearable data, compulsory advertisement), IV (fake news indistinguishable from reality, reverse Turing Tests), and V (ad-hoc social networks, cryptojacking for the human brain).

Industry
Why you should not use Google Cloud. (Medium)
“This post is not about the quality of Google Cloud products. They are excellent, on par with AWS. This is about the “no-warnings-given, abrupt way” they pull the plug on your entire systems if they (or the machines) believe something is wrong.”

Apple Engineers Its Own Downfall With the Macbook Pro Keyboard (iFixit)
“A titan of tech and industrial innovation has been laid low by a mere speck of dust. Last week, Apple quietly announced that they were extending the warranty on their flagship laptop’s keyboard to four years. As it turns out, the initial run of these keyboards... has been magnificently prone to failure.”

Software
Free Monads Are Simple (Underscore)
In a follow-up from last week’s articles on Monads, Noel Walsh is here to show you that they’re simple: “the free monad brings together monads and interpreters...Interpreters are about separating the representation of a computation from the way it is run.”

AI/Machine Learning
OpenAI Five (OpenAI)
“Our team of five neural networks, OpenAI Five, has started to defeat amateur human teams at Dota 2… we aim to beat a team of top professionals at The International in August.”

Layoffs at Watson Health Reveal IBM’s Problem With AI (IEEE Spectrum)
“The company has touted Watson, its flagship artificial intelligence, as the premier product for turning our data-rich but disorganized world into a smart and tidy planet….But according to engineers swept up in a major round of layoffs within IBM’s Watson division last month, the company’s promotions of its “cognitive computing” platform mask its own real difficulties in turning its AI into a profitable business.”

Misc
The Control Group Is Out Of Control (Slate Star Codex)
If we wanted to test the effectiveness of the modern scientific process, we could use parapsychology (the study of psychic/paranormal phenomena) as the control group. “The results are pretty dismal. Parapsychologists are able to produce experimental evidence for psychic phenomena about as easily as normal scientists are able to produce such evidence for normal, non-psychic phenomena.”

Sleep Scientist Warns Against Walking Through Life 'In An Underslept State' (NPR)
Short summary of advice for better sleep/rest (hint: you’re not getting enough). And surprise, surprise: even though they can feel like they don’t harm you, alcohol, sleeping pills such as Ambien, and caffeine all lead to lower sleep quality, even if you don’t notice.

A Lost Secret: How To Get Kids To Pay Attention (NPR)
“Many studies have shown that when teachers foster autonomy, it stimulates kids' motivation to learn, tackle challenges and pay attention…[but the growing] lack of autonomy in school inhibits kids' ability to pay attention.”

Want to get in touch? Reply here, or visit nicholasbradford.io
Views expressed here are my own and not representative of any past or present affiliated entity.

6/25/2018 Tech Ingestor: The Post-Cubicle Office, The Free Monad, and Neural Rendering

Tech Ingestor

Best

The Post-Cubicle Office and Its Discontents (NYT)
“The move to take people out of private offices, the better to improve collaboration and productivity, has little empirical justification... A study from The Journal of Environmental Psychology in 2013 indicated that 50 percent of workers in open-plan spaces suffer from a lack of sound privacy, and 30 percent complain about a lack of visual privacy.”

Industry

Augur: A Decentralized Oracle & Prediction Market Platform (Augur)
"Augur allows users to trustlessly create prediction markets on the outcome of any future event", by providing financial incentives to holders of Reputation tokens.

Software

Why the free Monad isn't free - by Kelley Robinson (YouTube)
A very simple and intuitive explanation of what Monads are, why they’re useful, and why using them in your codebase can have hidden costs. To see just how complex things can become, see “True Scala Complexity” (yaaang's blog) - and for the opposite side of the story, see “Functional Patterns in Domain Modeling” (YouTube). For someone more neutral, also check out “Scala with Style” by Martin Odersky (YouTube).

AI/Machine Learning

Neural scene representation and rendering (DeepMind)
"We introduce the Generative Query Network (GQN)... composed of two parts: a representation network and a generation network. The representation network takes the agent's observations as its input and produces a representation (a vector) which describes the underlying scene. The generation network then predicts (‘imagines’) the scene from a previously unobserved viewpoint."

Model Tuning and the Bias-Variance Tradeoff (r2d3)
Provides an extremely well-crafted animated visual introduction to one of the fundamental trade-offs in building machine learning models.

Misc

I Can Tolerate Anything Except The Outgroup (Slate Star Codex)
Scott Alexander explores trends in tolerance and the right/left political divide: “Research suggests Blue Tribe / Red Tribe prejudice to be much stronger than better-known types of prejudice like racism.”

Where to Watch the World Cup in NYC (NYT)
"So what if the United States didn’t qualify for the 2018 World Cup? In New York you can eat, drink and watch the games with Serbs, Swedes, Senegalese or wherever there’s a TV over the bar."

Want to get in touch? Reply here, or visit nicholasbradford.io
Views expressed here are my own and not representative of any past or present affiliated entity.

6/18/2018 Tech Ingestor: Internet trends, DeepFakes, and Ranting about Platforms

Tech Ingestor

Best

Mary Meeker’s 2018 internet trends report: All the slides, plus analysis (Recode)
“It’s that time of year again, when Mary Meeker unloads her highly anticipated internet trends report for the Code Conference crowd in Rancho Palos Verdes, Calif. This year, the Kleiner Perkins Caufield & Byers partner released 294 slides in rapid succession, covering everything from smartphone behavior in the U.S. to tech company competition in China.”

Industry

The Robots Are Coming for Wall Street (NYT)
Within 10 years, “between a third and a half of the current employees in finance will lose their jobs to Kensho and other automation software. It began with the lower-paid clerks... It has moved on to research and analysis, as software like Kensho has become capable of parsing enormous data sets far more quickly and reliably than humans ever could. The next ‘‘tranche’’... will come from the employees who deal with clients.”

Chatbots were the next big thing: what happened? (GrowthBot)
“At the Mobile World Congress 2017… the only significant question around chatbots was who would monopolize the field, not whether chatbots would take off in the first place. One year on, we have an answer to that question. No. Because there isn’t even an ecosystem for a platform to dominate.”

Software

Stevey's Google Platforms Rant; from 2011 (Google+)
An older but significant blog post from Steve Yegge about the contrasting approaches to building platforms at Amazon vs Google. Incidentally, Yegge recently left Google to join Grab and wrote extensively about the decision on Medium.

Interviewing is a noisy prediction problem (Erik Bernhardsson’s blog)
The CTO of Better and a conductor of ~2000 interviews gives his thoughts. “The correlation between who did really well in the interview process and who performs really well at work is really weak. Confronted by this observation I’ve started thinking about this process as inherent noise reduction problem.”

AI/Machine Learning

The US military is funding an effort to catch deepfakes and other AI trickery (MIT Tech Review)
“Using what are known as generative adversarial networks, or GANs, it is possible to generate stunningly realistic artificial imagery… ‘Theoretically, if you gave a GAN all the techniques we know to detect it, it could pass all of those techniques...We don’t know if there’s a limit. It’s unclear.’”

AI at Google: our principles (Google blog)
Sundar Pichai outlines 7 objectives, along with with a set of AI applications that Google will not pursue (see Google won’t renew it’s DoD contract for Project Maven).

Realtime tSNE Visualizations with TensorFlow.js (Google AI Blog)
“With this implementation, what used to take 15 minutes to calculate (on the MNIST dataset) can now be visualized in real-time and in the web browser. Furthermore this allows real-time visualizations of much larger datasets, a feature that is particularly useful when deep neural output is analyzed.”

Want to get in touch? Reply here, or visit nicholasbradford.io
Views expressed here are my own and not representative of any past or present affiliated entity.

6/4/2018 Tech Ingestor: Invisible Asymptotes, AI Winter, and Interviewing Evan Spiegel

Tech Ingestor

Best

Invisible Asymptotes (Eugene Wei’s Blog)
“In strategic planning [at Amazon] the question in building my forecast was to flush out what I call the invisible asymptote: a ceiling that our growth curve would bump its head against if we continued down our current path. It's an important concept to understand for many people in a company, whether a CEO, a product person, or, as I was back then, a planner in finance.” A longer read with several tech company case studies, but worth it.

Industry

Snap CEO Evan Spiegel | Full interview | 2018 Code Conference (YouTube)
A 40-minute interview on what separates Snap from other social media, diversity, redesign mistakes, and more. On Facebook copying Snap: “We would really appreciate it if they copied our data protection practices”.

Holacracy and the mirage of the boss-less workplace: Lessons from the failures at Github, Medium & Buffer (Medium)
The “innovative company organization system that favors flat structures over a classical hierarchy, was all the rage a few years ago… but over the last few months, these companies began renouncing Holacracy and returning to a more conventional pyramid-shaped command structure.”

Software

Why is Front-End Development So Unstable? (Jimmy Breck-McKye’s Blog)
Quantifying the issues with our existing paradigm (i.e. NPM), a walkthrough of a junior dev’s headaches, and potential mitigation steps.

AI/Machine Learning

AI Winter Is Well On Its Way (Piekniewski's blog)
“Deep learning has been at the forefront of the so called AI revolution for quite a few years now, and many people had believed that it is the silver bullet that will take us to the world of wonders of technological singularity...But this narrative begins to crack.”

Eigenvectors and Eigenvalues explained visually (Setosa)
“Eigenvalues/vectors are instrumental to understanding electrical circuits, mechanical systems, ecology and even Google's PageRank algorithm. Let's see if visualization can make these ideas more intuitive.”

Misc

Meditations on Moloch (Slate Star Codex)
“In some competition optimizing for X, the opportunity arises to throw some other value under the bus for improved X. Those who take it prosper. Those who don’t take it die out. Eventually, everyone’s relative status is about the same as before, but everyone’s absolute status is worse than before.”

The financial scandal no one is talking about (The Guardian)
“Accountancy used to be boring – and safe. But today it’s neither. Have the ‘big four’ firms become too cosy with the system they’re supposed to be keeping in check?”

Managed by Q’s ‘Good Jobs’ Gamble (NYT)
“Forgoing the gig-economy model, a start-up bets on a strategy that puts cleaning-service workers on a professional path.” Investing more in employees limits short-term process, but creates a much more stable customer and employee base.

A Half-Day Of Diversity Training Won’t Change Much For Starbucks (FiveThirtyEight)
“Diversity training in general doesn’t change much for any corporation.”

Want to get in touch? Reply here, or visit nicholasbradford.io
Views expressed here are my own and not representative of any past or present affiliated entity.

5/29/2018 Tech Ingestor: Sapiens, Stuxnet, and The Boring Company

Tech Ingestor

Best

Sapiens: A Brief History of Humankind, summary and review (LessWrong)
This sweeping overview of the last 130,000 years comes with recommendations from Bill Gates, Mark Zuckerberg, and Barack Obama. Inside are intriguing ideas including: that humans actually grew less happy after the Agricultural Revolution, the power of collective fictions (e.g. money, animism, corporations, liberalism) as great unifiers, and the potential for Sapiens as we know it to vanish in the wake of scientific advances.

Software: Industry

How I targeted the Reddit CEO with Facebook ads to get an interview at Reddit (Twicsy’s Blog)
“Steve Huffman, CEO of Reddit, saw my ad, clicked on it, read (probably skimmed) my article, and liked it well enough to send a note to Reddit HR to contact me about a position.”

Breaking a myth: Data shows you don’t actually need a co-founder (TechCrunch)
Data from CrunchBase contradicts the common advice to always find a co-founder. Caveat: there’s still some strong anecdotal evidence of prominent investors who are less likely to fund a solo-founder startup.

Software

What is the most sophisticated piece of software/code ever written? (Quora)
A look inside Stuxnet, the (allegedly American/Israeli) computer worm used to disrupt the uranium enrichment centrifuges used by Iran’s nuclear program.

AI/Machine Learning

Wall Street Tech Spree: With Kensho Acquisition S&P Global Makes Largest A.I. Deal In History (Forbes)
Kensho was sold to S&P Global for $500mm. It’s mission is to use machine learning to make complex financial analysis as easy as a search on Google, presenting a plain English text box interface for questions, and drawing upon a huge historical corpus and custom “cross correlation engine”.

AI and Compute (OpenAI Blog)
“Since 2012, the amount of compute used in the largest AI training runs has been increasing exponentially with a 3.5 month-doubling time (by comparison, Moore’s Law had an 18-month doubling period).”

Smart Compose: Using Neural Networks to Help Write Emails (Google AI Blog)
Introduced at Google I/O as a new Gmail feature which offers sentence completion suggestions as you type, for faster email drafting. Design considerations included latency, scale, and interestingly, the elimination of human cognitive biases permeating the training set.

Google Duplex: An AI System for Accomplishing Real-World Tasks Over the Phone (Google AI Blog)
More details into the system demoed recently at Google I/O. An important caveat: “Duplex can only carry out natural conversations after being deeply trained in [closed domains, which are narrow enough to explore extensively]. It cannot carry out general conversations.”

Misc

The Boring Company: Information Session (YouTube)
An intriguing presentation by Musk on the company’s vision (for alleviating city traffic with tunnel networks), including: how they’re planning to achieve an order of magnitude improvement in tunnel-boring machines, why earthquakes aren’t cause for concern, and a Hyperloop prototype for inter-city travel faster than by aircraft.

The Cult of the Root Cause (Reinertsen & Associates)
We often think that asking “why” continuously to get to the root cause is the best way to approach problems, but “Don’t assume you will only encounter problems that can be reduced to a simple single chain of causality where the best intervention lies at the start of the chain. Be open to the possibility that you are dealing with a causal network that has multiple starting points and endpoints.”

Want to get in touch? Reply here, or visit nicholasbradford.io
Views expressed here are my own and not representative of any past or present affiliated entity.

5/21/2018 Tech Ingestor: Work+Life Algorithms, Google’s new tools, and re-writing your code base “from scratch”

Tech Ingestor

Best

a16z Podcast: Principles and Algorithms for Work and Life (a16z)
Ray Dalio sits in with VC firm Andreessen Horowitz to discuss “everything from the differences between private and public investing, and between startups and big companies — to how people, teams, organizations, and even nation-states can evolve through principles like “believability-weighted idea meritocracies” and more.”

Software: Industry

Google’s new tools will make your life more convenient—for a price (MIT Tech Review)
Some highlights from Google I/O: Duplex (human-sounding AI agent that can set up appointments for you), Smart Compose for GMail (suggestions for phrases as you type), Style Match (searches for similar shopping items directly from the Camera), and Dashboard (for tracking usage).

Software

Things You Should Never Do, Part I (Joel on Software)
Trying to rewrite a whole code-base is a massive strategic mistake, and takes a significant amount of arrogance to presume both that you’ll be able to re-create all the tiny bug fixes and edge cases that made the first version “bad”, and that you’ll be able to write it better than you did the first time.

Pyre: Fast Type Checking for Python (Facebook)
Pyre is designed to help improve the quality and development speed in large Python codebases by flagging type errors interactively in your favorite editor.

AI/Machine Learning

One model to learn them all (The Morning Paper)
We demonstrate, for the first time, that a single deep learning model can jointly learn a number of large-scale tasks from multiple domains. The key to success comes from designing a multi-modal architecture in which as many parameters as possible are shared and from using computational blocks from different domains together.

How to solve 90% of NLP problems: a step-by-step guide (Insight Data Science blog)
Walks through an example problem (“classify a tweet as about a natural disaster or not”) with gradually more sophisticated approaches, with particular attention paid to examining and explaining model behavior.

How I Fail – Ian Goodfellow (PhD’14, Computer Science) (Veronikach)
A staff research scientist on the Google Brain team, who was included in MIT Technology Review’s “35 under 35” as inventor of generative adversarial networks, shares his failures.

Misc

Shortware Trading Part 1: The West Chicago Tower Mystery (Sniper in Mahwah)
An intriguing investigation into a particular radio tower linked to the CME, as part of a wireless network used by HFT firms to reduce latencies between exchanges around the world.

‘I had to guard an empty room’: the rise of the pointless job (The Guardian)
“Copying and pasting emails. Inventing meaningless tasks for others. Just looking busy. Why do so many people feel their work is completely unnecessary?”

Want to get in touch? Reply here, or visit nicholasbradford.io
Views expressed here are my own and not representative of any past or present affiliated entity.

5/7/2018 Tech Ingestor: The man of one study, large JS applications, and the lack of an AI Revolution

Tech Ingestor

Best

Beware the Man of One Study (Slate Star Codex)
“At some point in their education, most smart people usually learn not to credit arguments from authority. If they’re really smart, they’ll use the magic words “peer-reviewed experimental studies.” But I worry that most smart people have not learned that a list of dozens of studies, several meta-analyses, hundreds of experts, and expert surveys showing almost all academics support your thesis – can still be [wrong].”
Software

Designing very large (JavaScript) applications (Medium)
By a staff engineer on JS infrastructure at Google, and also creator of the AMP Project. “The best advice I can give: Don’t let your applications get very large. The best way to not get there is to delete stuff before it is too late.”

Why it took a long time to build that tiny link preview on Wikipedia (Wikimedia blog)
It took several years to complete this feature, due to tricky details such as: which thumbnail to pick, how to summarize the article, which HTML elements are relevant, feedback rounds, and scaling the preview API (now receiving over 500mm hits/minute).

AI/Machine Learning

Artificial Intelligence — The Revolution Hasn’t Happened Yet (Medium)
Michael I. Jordan (extremely influential AI researcher) outlines the larger challenges faced the emerging discipline of AI, and suggests how to refocus - including that while we might use “AI” as a nebulous umbrella term, there is critical work to be done in Intelligence Augmentation and Intelligence Infrastructure. “This is not the classical case of the public not understanding the scientists — here the scientists are often as befuddled as the public.”

The surprising creativity of digital evolution (The Morning Paper)
A collection of amusing anecdotes from the evolutionary computation community, as often it turns out the best way for an evolutionary algorithm to succeed is to find a loophole in the reward function.

The Fraudulent Claims Made by IBM About Watson and AI (Roger Schank’s blog)
“It would be nice if IBM would tone down the hype and let people know what Watson can actually do and stop making up nonsense about love fading and out thinking cancer.”

Lessons from My First Two Years of AI Research (Tom Silver’s Blog)
Finding good mentors, investing in useful skills, finding and digesting papers, and making steady progress on long-term research.

Misc

9 Tricks to Appear Smart in Brainstorming Meetings (Medium)
A very amusing read - “Nothing makes you seem smarter than when you question the questions by asking if they’re the right questions. If someone responds by asking you what you think the right questions are, say you just asked one.”

China is mining data directly from workers’ brains on an industrial scale (SCMP)
Government-backed surveillance projects are deploying brain-reading technology to detect changes in emotional states in employees on the production line, the military and at the helm of high-speed trains

Can’t sleep? Tell yourself it’s not a big deal (The Guardian)
There’s growing evidence that thinking of yourself as an insomniac is a major part of the problem.

Is there a Fix for Impostor Syndrome? (ACM Interactions)
Feelings of inadequacy/incompetence are common, but there are environmental changes that can help mitigate this. Also helpfully breaks down Impostor Syndrome archetypes into the Expert, the Perfectionist, the Superwoman/man, the Natural Genius, and the Rugged Individualist.

Goldman Sachs to Open a Bitcoin Trading Operation (NYT)
“They’re moving ahead with plans to set up what appears to be the first Bitcoin trading operation at a Wall Street bank... Goldman will begin using its own money to trade Bitcoin futures contracts on behalf of clients.”

Want to get in touch? Reply here, or visit nicholasbradford.io
Views expressed here are my own and not representative of any past or present affiliated entity.

Tech Ingestor: Banning Telegram, Keeping Palantir, and the Failure to Lunch

Tech Ingestor
Best

Failure to Lunch: The lamentable rise of desktop dining (NYT)
“Now some 62 percent of professionals say they typically eat lunch at their desks,” but “workplace satisfaction is so much higher if you eat with your colleagues”.

Software: Industry

Russia’s Telegram ban is a big, convoluted mess (The Verge)
After Russia banned Telegram when it refused to hand over encryption keys (“terrorists can use encrypted messaging services to organize”), local ISPs started tried blocking Telegram by blocking 15.8 million IPs on Amazon and Google’s cloud platforms. Big surprise, that made a lot of other essential web services offline.

Palantir Knows Everything About You (Bloomberg)
“Peter Thiel’s data-mining company is using War on Terror tools to track American citizens”. Not only is this susceptible to immoral use by operators, but what about when the underlying data/algorithms are flawed?

Here’s how hackers could cause chaos in this year’s US midterm elections (MIT Tech Review)
“Despite efforts to boost security, critical parts of America’s voting infrastructure are still vulnerable to cyberattack.” I highly recommend “Why Electronic Voting is a Bad Idea” by Computerphile as a supplement.

The Stack Overflow Age (Joel on Software)
The CEO and co-founder describes the Q&A ecosystem during the early days of the Internet, and what made Stack Overflow (started in 2008) different.

Software

Software Testing Anti-patterns (Codepipes Blog)
A list of 13 anti-patterns, their sources, and solutions - including favorites such as “Paying excessive attention to test coverage”, “Not converting production bugs to tests”, and “Testing the wrong functionality”.

The architectural implications of autonomous driving: constraints and acceleration (The Morning Paper)
A walkthrough of the design considerations (avg. latency, tail latency, energy usage) when complementing CPUs with GPUs, FPGAs, and ASICs and running state-of-the-art algorithms (for object detection, tracking, localization, fusion, and motion planning).

AI/Machine Learning

Using Evolutionary AutoML to Discover Neural Network Architectures (Google Research Blog)
Allowing simple evolutionary algorithms (running on days over hundreds of TPUs) to build network architectures from simple building blocks (some expert opinion goes into picking those blocks) generates results that match or exceed hand-tuned architectures. “Evolution is faster than reinforcement learning in the earlier stages of the search… Moreover evolution is quite robust to changes in the dataset or search space."

Personalized Hey Siri (Apple ML Journal)
Apple introduced the always-on “Hey Siri” feature with the iPhone 6S, which comes with the challenge of limiting false positives introduced when a person other than the phone’s owner says “Hey Siri”. This blog covers the motivation, general problem, user enrollment workflow, system overview, and evaluation of Apple’s technique for tackling this problem.

A.I. Researchers Are Making More Than $1 Million, Even at a Nonprofit (NYT)
“OpenAI paid its top researcher, Ilya Sutskever, more than $1.9 million in 2016. It paid another leading researcher, Ian Goodfellow, more than $800,000.”

Misc

Teen charged in Nova Scotia government breach says he had 'no malicious intent' (CBC)
“The teen has been charged with "unauthorized use of a computer," which carries a possible 10-year prison sentence, for downloading approximately 7,000 freedom-of-information releases… the 19-year-old says he thought he was downloading an archive of public information that was supposed to be freely available on the internet.”

Robots Continue Attempting to Master Ikea Furniture Assembly (IEEE Spectrum)
“We’ve seen a variety of robotic systems tackle Ikea in the past, but today… a mostly off-the-shelf system of a few arms and basic sensors that can put together the frame of a Stefan chair kit autonomously(ish) and from scratch.”

Suggestions for improvements I can make? Want to get in touch? Reply here, or visit nicholasbradford.io
Views expressed here are my own and not representative of any past or present affiliated entity.

Tech Ingestor: Productivity, mind-reading, and Tesla replacing robots with humans

You’ve probably already read about Zuckerberg’s performance in front of lawmakers, and the resulting implications about our data - so here are some memes about it.

Best

Productivity (Sam Altman’s Blog)
Compounding effects, what you work on, prioritization, physical factors, and misc advice. “The right goal is to allocate your year optimally, not your day.” Also check out closely-related “How to Decide What to Build” by Daniel Gross.

Researchers develop device that can 'hear' your internal voice (The Guardian)
MIT’s AlterEgo headset can transcribe words that wearers verbalise internally using only electrodes attached to the skin.

China Now Has the Most Valuable AI Startup in the World (Bloomberg)
SenseTime specializes in large-scale facial recognition software, of the kind recently used to catch a man at a crowded pop concert. It has raised $600 million from Alibaba and other investors at a valuation of over $3 billion.

Software: Industry

In a Leaked Memo, Apple Warns Employees to Stop Leaking Information (Bloomberg)
Apple caught 29 “leakers” last year, and noted that 12 of those were arrested. “Leakers do not simply lose their jobs at Apple. In some cases, they face jail time and massive fines for network intrusion and theft of trade secrets both classified as federal crimes.”

Three years of the Right To Be Forgotten (The Morning Paper)
2.4M URLs have been requested for delisting from the Google search engine over the last 3 and a half years. “Critically, the [RTBF] ruling requires that search engine operators make the determination for whether an individual’s right to privacy outweighs the public’s right to access lawful information when delisting URLs.”

The YC Seed Deck Template (YCombinator Blog)
“The key point to remember here is that founders should strive for clarity and concision. This is not the right place to write a treatise on your market or world philosophy.”

Software

Being Popular (Paul Graham’s Blog)
Coming from a “hacker” lens, ideas on what makes a programming language popular, including: Brevity, Hackability, External Factors, Libraries, Efficiency, and more. Also see his related ideas on future languages in The Hundred-Year Language, and thoughts on why Java specifically “smells suspicious” (in terms of origin/spread, not technical properties) in Java’s Cover.

Why Does "=" Mean Assignment? (Hillel Wayne’s Blog)
“Looking at this as a whole, = was never “the natural choice” the assignment operator… Nowadays most languages use = entirely because C uses it, and we can trace C using it to CPL being such a trash fire.”

AI/Machine Learning

Commoditisation of AI, digital forgery and the end of trust: how we can fix it (Giorgio Patrini’s blog)
“It is becoming widely evident that technology will enable total manipulation of video and audio content, as well as its digital creation from scratch. As a consequence, the meaning of evidence and trust will be critically challenged.”

Interpreting predictive models with Skater: Unboxing model opacity (O’Reilly Data Science)
A deep dive into model interpretation as a theoretical concept and a high-level overview of Skater, which can provide interpretations at both the dataset and single-prediction level.

What Happens When an Algorithm Helps Write Science Fiction (Wired)
University of Toronto researchers are developing software (“SciFiQ”) which can identify dozens of structural and stylistic details in huge chunks of text, and so can extract similarities in great works. One writer used this to try to assist him in writing the perfect algorithm-assisted short story.

Crypto

A survey on security and privacy issues of Bitcoin (The Morning Paper)
A helpful catalogue of the major types of attacks on a cryptocurrency network, with known mitigations.

Misc

Turns out there's a simple fix for Model 3 production issues: Fewer robots (Mashable)
In an interview with CBS Good Morning, Musk explains that the excessive automation is actually slowing output - and also reveals that he’s been spending his nights in a sleeping bag on a conference room floor.

A Lifesaving Checklist (NYT)
Enforcement of simple checklists can have drastic effects when humans are expected to repeatedly perform mundane tasks without error, as with a 5-step infection prevention checklist for Michigan ICUs that saved 1,500 lives over 18 months. Atul Gawande later wrote “The Checklist Manifesto” on this subject, and you can read the first chapter here.

This is the most dexterous robot ever created (MIT Tech Review)
“The most nimble-fingered machine yet shows how machine learning can teach robots to recognize and pick up different types of objects, a skill that could transform many factories and warehouses.”

Microplastics found in 93% of bottled water tested in global study (CBC)
“Researchers examined 11 different brands of water purchased in 9 countries” - although the brands in question contest that the plastic particle density is much lower than these researchers claimed (average of 10.4 particles/litre).

Credit Card Signatures Are About to Become Extinct in the U.S. (NYT)
Credit card companies have been trying to get rid of them for years, and now they’re finally succeeding.

Views expressed here are my own and not representative of any past or present affiliated entity.

Tech Digest 4/6/2018: Blind hiring, biological clocks, and dark patterns

Approved to send home, as always. Stay tuned for the next installment, in which you’ll discover a way to get this email delivered to your personal email address directly.

Best

Is Blind Hiring the Best Hiring? (NYT, 2016): “Most companies say they want to attract a diverse workforce, but few deliver.” Anonymity during the interview process may be a solution.

Poor grades tied to class times that don’t match our biological clocks (Berkeley News, 2018): After sorting nearly 15,000 students into “night owls,” “daytime finches” and “morning larks”, researchers compared their class times to their academic outcomes, and found that mismatches between biological rhythm and class time indeed negatively impacted grades.

AI/Machine Learning

Adversarial Patch (The Morning Paper, 3/29/2018): Details a way to way to create funky printable stickers which trick image classifiers into identifying completely wrong objects (e.g., placing the “toaster” pattern sticker next to a banana on a table causes the classifier to go from identifying “banana” with 97% confidence to “toaster” with 99% confidence).

Are dogs appreciated fairly on @dog_rates? (Vince Knight’s Blog, 2018): A statistical investigation to confirm that the number of likes and retweets a dog gets on @dog_rates isn’t influenced by the rating - because after all, they’re all good dogs.

Comet.ml (Comet.ml): Web platform which “allows data science teams and individuals to automagically track their datasets, code changes, experimentation history and production models creating efficiency, transparency, and reproducibility.”

Software

Using Services to Break Down Monoliths (Yelp Engineering, 2015): Yelp went from a single monolith in 2012 to 70+ services in 2015, allowing for a highly scalable dev/test/deploy process. They’ve also outlined their own architecture principles for easy consumption.

Effective Engineer – Notes (GitHub, 2018): A terse summary of Edmond Lau’s book, focusing on leveraging yourself, optimizing for learning, prioritization, investment in the team and more.

Software - Industry

LinkedIn Dark Patterns or: Why Your Friends Keep Spamming You to Sign Up for LinkedIn (Medium, 2015): Their address book import design is harmful to their brand, even if it’s better for business in the short term.

How Dark Patterns Trick You Online, by The Nerdwriter (YouTube, March 2018): Builds off the ideas of “LinkedIn Dark Patterns” with further examples (e.g. it’s impossibly difficult to delete your Amazon account) and their implications.

Fake news 2.0: personalized, optimized, and even harder to stop (MIT Tech Review, 3/27/2018): “Artificial intelligence will automate and optimize fake news, warns a technology supplier to US intelligence agencies.”

Apple Plans to Use Its Own Chips in Macs From 2020, Replacing Intel (Bloomberg, 4/2/2018): They make up 5% of Intel’s revenue. Apple is also reported working on software platform to merge iOS and macOS apps.

Thousands of Google Employees Protest Company's Involvement in Pentagon AI Drone Program (Gizmodo, 4/4/2018): Google’s involvement with Project Maven, intended to help the Pentagon use artificial intelligence to analyze drone footage, has drawn ire from the Google’s employees, and thousands have now signed a petition urging Sundar Pichai to shut it down.

Crypto

How network theory predicts the value of Bitcoin (MIT Tech Review, 3/29/2018): “Metcalfe’s Law, which measures the value of a network, can calculate a cryptocurrency’s value—and predict when to get out.”

Tech Digest 3/23/2018: Zuckerberg’s Reckoning

Best

Mark Zuckerberg’s Reckoning: ‘This Is a Major Trust Issue’ (NYT, 3/21/2018): After the recent Cambridge Analytica data privacy debacle (and a nearly 10% drop in Facebook’s stock price, along with #DeleteFacebook trending), Mark finally addresses the concerns.

Meet is Murder (NYT, 2016): Are people fundamentally divided into Makers and Managers? Can Slack replace most of our meetings? And would adopting “Robert’s Rules” help us run our meetings more efficiently? (Hint: yes). Paul Graham’s essay on the subject is also worth reading. “Don't your spirits rise at the thought of having an entire day free to work, with no appointments at all? Well, that means your spirits are correspondingly depressed when you don't.”

Software - Industry

Stack Overflow Developer Survey 2018 (Stack Overflow, 2018): Demographics, distribution of experience, favorite frameworks, and life habits - all of it’s here. (Scala might not be up-and-coming like Python, but it’s the second highest-paid language in the US after Erlang).

A startup is pitching a mind-uploading service that is “100 percent fatal” (MIT Tech Review, 3/13/2018): YCombinator’s Nectome will nearly perfectly preserve your brain with the assumption that we will crack brain-uploading before it degrades - but the process only works on “fresh brains”, so you have to be euthanized in order to use it.

Misc

Pixar’s 22 Rules of Storytelling (Aerogramme, 2013): “Pixar’s Rules of Storytelling were originally tweeted by Emma Coats, Pixar’s Story Artist.”

China to bar people with bad 'social credit' from planes, trains (Reuters, 2/16/2018): “The move is in line with President’s Xi Jinping’s plan to construct a social credit system based on the principle of ‘once untrustworthy, always restricted’”. The number of people banned from flights is already over 6 million.

Machine Learning

Your next computer could improve with age (MIT Tech Review, 2/12/2018): Google presents a preliminary scoping paper on how deep learning can prove to be a superior prefetching method.

When Machine Learning Matters (Erik Bernhardsson’s blog, 2016): “Most ML is sprinkles on top of the product” that’s been around for 5-10 years and needs another 10% performance/profit squeezed out of it.

Tech Digest 3/9/2018

Best

God Is a Bot, and Anthony Levandowski Is His Messenger (Wired, 9/27/2017): The insane story of the self-driving car guru’s instrumental role in building out Google’s capabilities (to the tune of over $100mm in personal compensation), and subsequent fall from grace as his shady techniques and move to Uber have led to a notorious IP lawsuit (Google and Uber settled in February). In the meantime he also founded an AI religion.

What Google Learned From Its Quest to Build the Perfect Team (NYT, 2016): Instead of relying on amassing an army of classically efficient and stoic individual contributors, the best teams are made of people who are comfortable with each other, feel free to share their thoughts and opinions, and contribute equally to group discussions. “We want to know that those people really hear us. We want to know that work is more than just labor.” (I’m trying hard not to use “Radical Truth and Transparency” and “Meaningful Work, Meaningful Relationships”, but...)

Finance

Risk Parity by Wealthfront Research (Wealthfront): An outside view and more math-heavy validation of Risk Parity, complete with Wealthfront’s own simulations compared to 60/40, Bridgewater’s All Weather, and AQR’s Risk Parity fund. (PS: did you know that Burton Malkiel, author of A Random Walk Down Wall Street, is Wealthfront’s CIO?)

Software - Industry

The Feds Can Now (Probably) Unlock Every iPhone Model In Existence (Forbes, 2/26/2018): Israel-based vendor Cellebrite “can retrieve (without needing to root or jailbreak the device) the full file system to recover downloaded emails, third-party application data, geolocation data and system logs” for all iOS versions up to 11.2.6 (the most recent).

Consent in the Digital Age: Can Apps Solve a Very Human Problem? (NYT, 3/2/2018): Straight from the “Hang the DJ” episode of Black Mirror, this app stores dynamic records of consent between users (on a blockchain, of course).

GitHub Survived the Biggest DDoS Attack Ever Recorded (Wired, 3/1/2018): At 1.35 terabits per second, it was even larger than the DDoS that took down Dyn in 2016.

YouTube's New Moderators Mistakenly Pull Right-Wing Channels (Bloomberg, 2/28/2018): “The misstep pulls YouTube, Google and parent Alphabet Inc. deeper into a toxic political fights over gun control, fake and extreme content, and whether internet companies should be responsible for what third parties post on their services.”

AI/Machine Learning

Reddit AMA with Google/Facebook/Microsoft AI Researchers (Reddit, 2/18/2018): Lots of interesting Q&A with Yann LeCun, Eric Horvitz, and Peter Norvig.

Why even a moth’s brain is smarter than an AI (MIT Tech Review, 2/19/2018): A neural network that simulates the way moths recognize odors also shows how they learn so much faster than machines.

Crypto

Venezuelans Can’t Buy Maduro’s Cryptocurrency With Bolivars (Bloomberg, 2/22/2018): You can only buy Petros with USD or Euro, and Venezuelans are prohibited from exchanging Bolivars for other currencies, so there’s not actually a good way for them to obtain their own country’s crypto.

46% of Last Year’s ICOs Have Failed Already (Bitcoin.com, 2/23/2018): It’s 59% including ICOs with teams that have stopped communicating on social media, or with an community of negligible size.

The world's first 100% honest Ethereum ICO (UEToken): The Useless Ethereum Token makes no claims to be useful for anything, and has raised $267,873.

Misc.

How poverty changes your mindset (Chicago Booth, 2/19/2018): “Working through a difficult financial problem produces a cognitive strain that’s equivalent to a 13-point deficit in IQ or a full night’s sleep lost...While scarcity can help people focus on costs and benefits, it can also cause stress that shifts attention and steals cognitive bandwidth.”

Why Is It Hard to Make Friends Over 30? (NYT, 2012): Turns out, lots of different reasons. “We joked about our inability to find time to hang out, and made a dinner date at the next available opening. It is three months from now.”

If you’re so smart, why aren’t you rich? Turns out it’s just chance. (MIT Tech Review, 3/1/2018): The most successful people are not the most talented, just the luckiest, a new computer model of wealth creation confirms. Taking that into account can maximize return on many kinds of investment.

Growing a company that sells miniature construction supplies to $17k a month (HackerNews): The founder of Mini Materials shared his journey, and the lessons he’s learned.

Tech Digest 2/23/2018: Facebook/Snapchat/Waymo aren't doing too hot

Best

Inside the Two Years that Shook Facebook - And the World (Wired, 2/12/2018): A long, detailed chronicle of Facebook’s evolving battle with free speech, moderation, Russians, and fake news.

Synthesizing Obama: Learning Lip Sync from Audio (UW Grail, 7/9/2017): Freakily realistic re-creation of mouth/face movements aligned to arbitrary speech.

Software

A sample of brilliance (The Morning Paper, 1/30/2018): Walks through Bob Floyd’s design of an efficient and beautiful random-sample algorithm.

Polynomial-time Algorithms for Prime Factorization on a Quantum Computer (The Morning Paper, 2/2/2018): An explanation of Shor’s Algorithm, the breakthrough work that showed it was possible to efficiently factor primes on a quantum computer (and the associated headaches for cryptography).

Facebook’s Zstandard - Real-time data compression algorithm (Facebook, 8/31/2016): Incorporates modern compression techniques to provide an improvement over Deflate (used in zip, gzip, …), and meant for general-purpose application over a variety of data types (check it out on GitHub).

Software-Industry

Facebook lost around 2.8 million U.S. users under 25 last year. 2018 won’t be much better. (Recode, 2/12/2018): According to eMarketer, it’s only going to get worse – although overall user growth will most likely remain positive, and Instagram remains strong.

In One Tweet, Kylie Jenner Wiped Out $1.3 Billion of Snap’s Market Value (Bloomberg, 2/22/2018): Regarding the [awful] recent UI update: “sooo does anyone else not open Snapchat anymore? Or is it just me... ugh this is so sad.”

After Settling With Uber, Waymo Faces Bigger Challenges (NYT, 2/10/2018): Competitors are catching up, Waymo still doesn’t have a business model, and talent is scarce (i.e. many of the best engineers are being snapped up by competing firms).

AI/Machine Learning

The Matrix Calculus You Need For Deep Learning (USFCA): “This paper is an attempt to explain all the matrix calculus you need in order to understand the training of deep neural networks. We assume no math knowledge beyond what you learned in calculus 1, and provide links to help you refresh the necessary math where needed.”

Why is machine learning in finance so hard? (Hardik Patel’s Blog, 2/11/2018): Changing distributions, small samples, unquantifiable data, complex linkages. “Financial time-series is a partial information game (POMDP) that’s really hard even for humans - we shouldn’t expect machines and algorithms to suddenly surpass human ability there.”

Spiking Neural Networks, the Next Generation of Machine Learning (Towards Data Science, 1/10/2018): SNN neurons use discrete “spikes” instead of outputting continuous values, have the local connectivity of a CNN, the built-in temporality of an RNN, and a closer resemblance to human neurons - too bad they aren’t differentiable and don’t work with current gradient-descent based optimization.

The “Black Mirror” scenarios that are leading some experts to call for more secrecy on AI (MIT Tech Review, 2/21/2018): “A new report by more than 20 researchers from the Universities of Oxford and Cambridge, OpenAI, and the Electronic Frontier Foundation warns that the same technology creates new opportunities for criminals, political operatives, and oppressive governments.” Included are thought-provoking dystopian scenarios than might not be so far off into the future.

Crypto

Salon asks ad-blocking users to opt into cryptocurrency mining instead (The Verge, 2/13/2018): Using Coinhive, which has been in the news recently for allowing malicious websites to mine crypto in-browser without your permission.

Bitcoin’s transaction fee crisis is over—for now (ARS Technica, 2/20/2018): “The median fee peaked at $34 in mid-December—now it's less than $1.” However, the reasons behind this aren’t particularly bullish for Bitcoin.

Misc.

The More Gender Equality, the Fewer Women in STEM (The Atlantic, 2/18/2018): “A new study explores a strange paradox: In countries that empower women, they are less likely to choose math and science professions.” Measured using the % of female STEM graduates VS. the World Economic Forum’s Global Gender Gap Index.

Crushed wood is stronger than steel (Nature, 2/7/2018): “Compressing the material and removing some of its polymers can increase its strength tenfold,” while previous densification methods only resulted in a 3-4x strength increase.

Tech Digest 2/8/2018: Re-creating images from brain scans, and Blockchain helps cobalt miners

Best

Blockchain to track Congo's cobalt from mine to mobile (Reuters, 2/2/2018): A pilot scheme to be launched this year should eventually give manufacturers a way of ensuring the cobalt in lithium-ion batteries has not been mined by children. Sealed bags of cobalt will receive a digital tag entered on the blockchain using a mobile phone (along with details of the weight, date, time and photo). Very interesting to read about a non-coin with big potential impact, as well as how many logistical barriers must be overcome for this plan to come to fruition.

Inside Amazon's Artificial Intelligence Flywheel (Wired, 2/1/2018): An intriguing chronicle of Amazon’s journey from very weak AI utilization and expertise to now using it to drive world-class innovation across the entire company. Perhaps an optimistic view of what our ML transformation could look like over the next few years?

Generating Images from Brain Activity (NVidia, 1/15/2018): Researchers in Japan developed a deep learning-based algorithm that can generate images from fMRI data, able to [very roughly] reconstruct simple shapes that the subject is viewing.

Life/Productivity

My New Favorite Book (GatesNotes, 1/26/2018) Enlightenment Now by Steven Pinker tracks a variety of metrics throughout history and applies it to 15 different measures of progress (like quality of life, knowledge, and safety), resulting in a holistic picture of how and why the world is getting better. Potential candidate for the IE Book Club?

GTD in 15 minutes – A Pragmatic Guide to Getting Things Done (Erlend Hamberg’s Blog): A simple task-management scheme, centered around the “In”, Next Actions, Waiting For, Project, and Someday lists - a summary of the popular book by David Allen.

Software

Ranking Programming Languages by GitHub Users (Ben Frederickson’s Blog,): Before you ask: Scala is at #17 with 0.78%. Rust, Kotlin, Go, and TypeScript are on the uptrend, as is Jupyter Notebook. PHP, Ruby, and Objective-C… not so much.

An intro to Probabilistic Programming with Ubers Pyro (YouTube-Siraj Raval, 11/10/2017): A <10min intro to methods for cleanly expressing and simulating over systems that depend on random variables. Disclaimer: Siraj’s wacky presentation style is not for everyone.

AI/Machine Learning

Fast.AI’s Practical Deep Learning for Coders 2018 (Fast.AI): A cutting-edge and intensely practical free online course that uses a “top-down, code-first” approach to get you up and running with the major concepts right away before going deep into the details. (Several ML club members are planning to take this over the next few months at a pace of roughly 1 lecture/2 weeks).

No Order Left Behind; No Shopper Left Idle (Instacart, 10/10/2017): To find the proper number of shoppers to keep at each store location at each hour, Instacart runs Monte Carlo simulations of orders and delivery routes to find the staffing schedule most likely to maximize profits (credit to Matt Willian for finding).

The UX of AI: Using Google Clips to understand how a human-centered design process elevates artificial intelligence (Google Design, 1/25/2018): Asking “what makes a photo memorable?” to guide the design and implementation of a camera which automatically recognizes and snapshots important moments. “The role of AI shouldn’t be to find the needle in the haystack for us, but to show us how much hay it can clear so we can better see the needle ourselves.”

This AI software dreams up new designs for 3-D-printed parts before your eyes (MIT Tech Review, 2/6/2018): I recommend taking 10 seconds to watch their demo animation, showing the design of a basic part evolve to satisfy a set of mechanical constraints with a funky and organic result.

Misc.

A search for insomnia genes involving 1.3 million people is the largest genetic study ever (MIT Tech Review, 2/2/2018): They found 956 different genes linked to the sleep disorder, but those still explain less than 10 percent of the overall chance that a person has it.

Chinese Police Add Facial-Recognition Glasses to Surveillance Arsenal (MIT Tech Review, 2/7/2018): sends data from its camera to a handheld device storing a 10,000-picture suspect database, crunching through in about 100 milliseconds to help officers spot criminals.

Banks in Britain and U.S. ban Bitcoin buying with credit cards (Reuters, 2/4/2018): Lloyds Banking Group and Virgin Money follow the lead of JPM and Citigroup. It’s apparently aimed at protecting consumers from funding purchases through credit card debt, and being unable to pay back if the value crashes.

1/26/2018

Best

Deep Learning: AlphaGo Zero Explained In One Picture (Applied Data Science, 1/4/2018): It learns purely from a “blank slate” with no human intervention or training data, using a Monte Carlo Tree Search (with a huge deep network to evaluate leaf nodes) to pick moves. This Medium post provides some additional context for why AlphaGo Zero is so significant.

Physicists create Star Wars-style 3D projections (Nature, 1/24/2018): Laser and particle system produces three-dimensional moving images that appear to float in thin air.

AI/Machine Learning

The case for learned index structures: Part 1 (The Morning Paper) and Part 2 (The Morning Paper): instead of the algorithms and data structures tuned over decades to optimize database lookups, why not train a neural network on your database to memorize the location mappings for constant-time lookup? You have to retrain the network every time, but allows for automatic workload personalization. “The current design, even without any significant modifications, is already useful to replace index structures as used in data warehouses, which might only be updated once a day.”

The AI Powered Equity ETF (NYSE: AIEQ) (EquBot ETF): A fund started in October that “seeks to provide investment results that exceed broad U.S. Equity benchmark indices at equivalent levels of volatility” by mysteriously using IBM Watson and slew of vague catchphrases.

Crypto/Blockchain

'Micro' Finance Giant Robinhood Makes Big Bet on Bitcoin Trading (CoinDesk, 1/25/2018): Robinhood plans to roll out commission-free bitcoin and ether trading services via its mobile apps next month (go to Robinhood’s site to get on the waitlist… oh wait we can’t use Robinhood).

Payment Processor Stripe to End Support for Bitcoin (CoinDesk, 1/23/2018): Citing lengthy transaction times, an increasing transaction failure rate, and growing fees causing bitcoin to become less popular among Stripe's merchants and users.

Misc.

Longevity FAQ: A beginner's guide to longevity research (Laura Deming’s Blog): The founder of the Longevity Fund (VC) sums up the active areas of research to extend human lifespans, including: caloric restriction, insulin-like growth factor, parabiosis (blood transfusions from young to old), senescence (eliminating old cells that have run out of telomerase), autophagy (improving lysosomes, the cellular “garbage disposal”), the hypothalamus (reducing core body temp causes it to release more growth hormones), reproductive systems (reducing fertility apparently can increase lifespans), mitochondria (tinkering with “oxidative stress”), and sirtuins (can change DNA folding/expression).

Harvard's milliDelta Robot Is Tiny and Scary Fast (IEEE Spectrum, 1/17/2018): A new kind of delta (parallel-motion) robot is only a few millimeters in size, but has incredible precision and has a payload twice its weight. Check out the videos of it tracing out geometric patterns at up to 75Hz.

Useful Mental Models (2016) (HackerNews,): a concise list of different mental models that may help with productivity, hypothesis testing, decision making, product design, and more.

CNN shutters Casey Neistat’s video company Beme, which it bought 14 months ago for $25 million (The Verge, 1/25/2018): Apparently due to “creative differences and sluggish process”. Neistat will also leave CNN.

1/19/2018

Weekly Tech Digest 1/19/18: Connecting Functional Programming to ML

Approved to send home (as always) – security issues should be cleared up, just double-check that you’re forwarding yourself a single email instead of a chain.

Also, what happens when your website depends on an NPM package that gets taken down without warning?

Best

The $25B eigenvector (2006) (HackerNews, 2018): “Google’s success derives in large part from its PageRank algorithm, which ranks the importance of webpages according to an eigenvector of a weighted link matrix. Analysis of the PageRank formula provides a wonderful applied topic for a linear algebra course.”

Neural Networks, Types, and Functional Programming (Chris Olah’s Blog, 2015): Each layer of a neural network is a link in a chain of pure functions, acting on the output of a previous layer and optimized to perform a specific task. Every layer also transforms the data into a new representation molding it into a form that makes their task easier to do - representations correspond to types. And using multiple copies of a neuron in different places is equivalent to sharing common functions (e.g. Recurrent Neural Nets are folds, Convolutional Neural Nets are maps with a sliding window). (Credit to Johnny Chang for finding).

Turning Design Mockups Into Code With Deep Learning (FloydHub Blog, 1/9/2018): Web designers could be getting their jobs automated away sooner than expected. This proof-of-concept ingests the raw design mockup image and whatever HTML it has generated so far, and produces the next HTML token to be appended to the output. Awesome because it includes “Mistakes I Made” asides and plenty of explanations/external links for learning.

Software

Many packages suddenly disappeared (HackerNews, 2018): A truly unsettling discussion about the potential build and security issues that occur when an open-source package included as a sub-dependency in many important libraries/projects suddenly disappear (e.g. taken down from NPM). This presents an opportunity for a malicious user to swoop in and publish a re-publish a package under the same name, but with malware added.

Who Y Combinator Companies Want (TripleByte, 12/9/2015): Programming interview site TripleByte shows that YCombinator startup interviews are basically filtering for clones of the founders, suffer from wild inconsistencies, bias strongly against deep technical or enterprise backgrounds in favor of those more product or design focused, and favor those with higher than junior-level experience (especially at one of the tech giants). (Aside: Max Howell, creator of Homebrew, has some strong feelings about programming interviews.)

Security

Meltdown (The Morning Paper, 1/15/2018): A more in-depth look with code examples. Meltdown loads a secure memory location into the register, uses a transient instruction (executed out-of-order and leaves behind measurable side effects) to access a cache line based on that contents, and uses the “Flush+Reload” side channel to access that cache line (and hence the secret). The KAISER patch going into OSs provides stronger kernel-user memory isolation.

Spectre attacks: exploiting speculative execution (The Morning Paper, 1/16/2018): A more in-depth look with code examples. With only a few lines of JavaScript and an address you want to read, can exploit the branch prediction and also use some Return-Oriented Programming techniques. It’s not as much one bug as an entire class of attacks that can be made against speculative execution in architectures.

Machine Learning

Generative Adversarial Networks (GANs) for Beginners (O'Reilly, 6/7/2017): Goes through the design and TensorFlow code for a simple GAN for generating handwritten digits from random noise, discriminated against by comparing to MNIST. Recommended to have some previous TensorFlow background.

Microsoft creates AI that can read a document and answer questions about it as well as a person (Microsoft AI Blog, 1/15/2018): Microsoft and Alibaba have models which beat human-level performance on the Stanford Question Answering Dataset (SQuAD - a dataset of 100,000+ question-answer pairs on 500+ Wikipedia articles). This is a major milestone on the road to building powerful text summarization and comprehension tools.

FinTech

Idea to Algorithm: The Full Workflow Behind Developing a Quantitative Trading Strategy, by Quantopian (YouTube, 12/7/2017): The crowd-sourced hedge fund demos its revamped workflow for doing research in IPython to generate signals, statistical tests to ensure validity, using their portfolio optimization tools to perform risk management, finding correlation to similar assets/algorithms, and testing on out-of-sample data (they also have a more detailed lecture series). Note - doesn’t seem like they’re doing great according to the Financial Times.

Best,
Nick

1/12/2018

Weekly Tech Digest 1/12/18: "Alchemy, Rigour, and Engineering" in AI

A week of review over the progress AI has made in 2017, and debating whether researchers are using proper rigor in their methods.

Best

AI and Deep Learning in 2017 – A Year in Review (WildML): A great overview of the past year, including AlphaGo, GANs, Neuroevolution, WaveNet, the “alchemy and engineering” debate, new hardware, and more.

Why Raspberry Pi isn’t vulnerable to Spectre or Meltdown (Raspberry Pi blog): Good additional explanation of the Spectre/Meltdown bugs, superscalar processors, branch prediction, and why Raspberry Pi's aren't vulnerable (credit to Brendan McManus for finding).

AI / Machine Learning

An overview of gradient descent optimization algorithms (Sebastian Ruder’s Blog): Compact summary of the modern variations on gradient descent (Adagrad, Adadelta, RMSprop, Adam, Nadam…), as well as additional techniques such as batch normalization and early stopping. Currently, Adam is probably the all-around best choice. Check out the Keras documentation for an additional concrete representation.

Welcoming the Era of Deep Neuroevolution (Uber AI Labs): Deep neural networks are trained routinely through stochastic gradient descent. However, Uber has released a set of five papers that support the emerging realization that neuroevolution, where neural networks are optimized through evolutionary algorithms (e.g. Genetic Algorithms), is also an effective method for reinforcement learning problems, and is poised to scale better in a future where we have increased access to massively parallelized computers.

Innovating Faster on Personalization Algorithms at Netflix Using Interleaving - Netflix Technology Blog (Medium): Describes how Interleaving, an optimization on A/B testing, can be used to used to find optimal video recommendation algorithms with user sample sizes 100x smaller than traditional methods. Disadvantages are that it take additional dev time to build support for, and is only useful for relative comparisons instead of absolute metrics.

Alchemy, Rigour and Engineering: A debate on the current state of ML

Taking Machine Learning from Alchemy to Electricity, by Ali Rahimi - NIPS 2017 Test-of-Time Award presentation (YouTube): Ali argues that Deep Learning research is overly focusing on experimental results without spending the requisite time developing theoretical building blocks.

Yann LeCun's (Facebook director of AI Research) rebuttal to Ali's talk (Facebook): “In the history of science and technology, the engineering artifacts have almost always preceded the theoretical understanding...Sticking to a set of methods just because you can do theory about it, while ignoring a set of methods that empirically work better just because you don't (yet) understand them theoretically is akin to looking for your lost car keys under the street light knowing you lost them someplace else.”

Alchemy, Rigour and Engineering (inFERENCe): It's okay to use non-rigorous methods, but it's not okay to use non-rigorous evaluation. You can think of "making a deep learning method work on a dataset" as a statistical test. “I would argue that the statistical power of experiments is very weak. We do a lot of things like early stopping, manual tweaking of hyperparameters, running multiple experiments and only reporting the best results. We probably all know we should not be doing these things when testing hypotheses. Yet, these practices are considered fine when reporting empirical results in ML papers.”

-Nick

1/06/2018

This week: an easy-to-read IPython blockchain implementation, details on the Meltdown and Spectre exploits, engaging in Deep Work, dating apps changing the fabric of society, and more.

Best

Dumbcoin - An educational python implementation of a bitcoin-like blockchain (GitHub): Really fascinating and easy-to-read IPython Notebook which allows you to understand the guts of simple cryptocurrency (lacks networking and a security guarantee).

Massive Intel CPU flaw: Understanding the technical details of Meltdown and Spectre (TechRepublic): Two critical architectural flaws in CPUs allow user processes to read kernel memory, affecting Intel, AMD, and ARM processors. Meltdown relates to the behavior of out-of-order execution on CPUs, neutralizing security models based on address space isolation, and is very easy to exploit, but has a software patch ready (at 10-30% CPU performance cost). Spectre is a flaw in branch prediction and speculation that allows applications to read kernel memory, and while more difficult to exploit may be hard to fix without hardware changes (and currently has no software patch).

Deep Work by Cal Newport | Book Summary (Paul Minor’s Blog): Covers: how so many people become busy without being productive, rewiring your brain to resist distractions by embracing boredom, giving yourself positive pastimes as an alternative to social media, and various tips and tricks for avoiding shallow work.

AI / Machine Learning

Deep Learning Scaling is Predictable, Empirically (Baidu Research): “Our empirical results show power-law generalization error scaling across a breadth of factors, resulting in power-law exponents---the "steepness" of the learning curve---yet to be explained by theoretical work. Further, model improvements only shift the error but do not appear to affect the power-law exponent.” Find the complete paper here.

When AI Supplies the Sound in Video Clips, Humans Can’t Tell the Difference (MIT Tech Review): Training models on thousands of 10-second YouTube clips in which the source of the sound is clearly visible in-frame, “... Evaluations show that over 70% of the generated sound from our models can fool humans into thinking that they are real”

Nobody's Ready for the Killer Robot (Bloomberg): A Q&A with General Robert Latiff on the ethics of warfare in the autonomous future. “I kind of view [the DoD’s approach] as a head-in-the-sand approach to the policies surrounding lethal autonomous weapons, and it cries out for some clarification.”

Blockchain

How Bitcoin Works Under the Hood (YouTube): The best explanation I’ve found so far for a blockchain beginner. Note: it’s from back in 2013.

Could Bitcoin technology help science? (Nature): Blockchain use could lead to “incorruptible data” - similar to several reports I’ve seen that banks may be able to significantly cut costs by switching their ledgers to blockchains.

Misc

Culture Queries: The perfect questions to ask in your job interview (Key-Value): A convenient interactive Map(cultural component that’s important to you -> best questions to ask during the interview)

First Evidence That Online Dating Is Changing the Nature of Society (MIT Tech Review): Dating websites have changed the way couples meet. Now evidence is emerging that this change is increasing levels of interracial marriage, and: “our model also predicts that marriages created in a society with online dating tend to be stronger”.

The science that’s never been cited (Nature): Nature investigates how many papers really end up without a single citation, which turns out to be a subtle and difficult question. While some often-repeated estimates are nearly 50%, it’s probably 21% or lower - and there are several reasons why uncited research can still be useful (e.g. when a paper closes off a future field of study).

How to read and organize online articles without driving yourself crazy (Gregory Ciotti’s Blog): The Regular Reads: use Feedly for organizing RSS feeds, and Unroll.Me for organizing all your newsletters into a single digest. One-off Articles: use Pocket for saving until later, Readability extension for Chrome when viewing on mobile, Delicious for organizing links with quick notes + tags, and Evernote for taking detailed notes.

12/22/2017

This week: Facebook releases research on potential negative effects of social media, Apple confirms that iPhones gradually slow their CPUs over time, Magic Leap AR releases a website, and blockchain remains a hot topic.

Highlights

Hard Questions: Is Spending Time on Social Media Bad for Us? (Facebook Research): The fact that Facebook is openly questioning the benefits and drawbacks of social media is really interesting. Summary: passively consuming information makes you feel measurably worse (possibly due to negative social comparison), but actively interacting with other people yields positive emotional results.

How to make your career plan (80,000 Hours): Given the “planning paradox” (the best option will keep changing as the world changes and you learn more), create a flexible plan such as the A/B/Z plan (by LinkedIn founder Reid Hoffman): your “ideal”, nearby alternatives, and your temporary fallback. Make sure to schedule self-check-ins every 6-12 months.

Tech

Magic Leap One (MagicLeap): The incredibly hyped augmented-reality startup (raised nearly $2bn) promises to ship in 2018, and has finally posted some intriguing features of their new platform for Creators. Check out the MIT Tech Review’s Analysis

Google Thinks I’m Dead, by Rachel Abrams (NYT): Requesting changes to errors in Google’s Knowledge Graph is nearly impossible, and can lead to embarrassing errors

Apple confirms iPhones with older batteries will take hits in performance (TheVerge): iPhone 6 and later received iOS updates which throttle CPU speed as battery degrades, sacrificing performance to improve the longevity of the electrical components. If your iPhone has gotten slow, replacing the battery will be much cheaper than buying a new phone

Amazon’s Fake Review Problem (Brian Bien’s Blog): This is a short and simple example, but the discussion comparing this to early exploitation of Google’s PageRank is interesting

Blockchain

Iced tea company says it's pivoting to the blockchain, and its stock jumps 200% (CNBC): Long Island Ice Tea changed name to Long Blockchain Corp after claiming to be in “preliminary stages” of evaluating potential business uses for blockchain

Statement on Cryptocurrencies and Initial Coin Offerings, by SEC Chairman Jay Clayton (SEC): Please, please be careful with cryptos. Also, the SEC is a little upset that not a single ICO tried to register with them, so plan on them not continuing to let that slide

Goldman Is Setting Up a Cryptocurrency Trading Desk (Bloomberg): should be up and running by the end of June

Miscellaneous

5 amazing books I read this year, by Bill Gates (GatesNotes)
- Evicted: Poverty and Profit in the American City, by Matthew Desmond. “If you want a good understanding of how the issues that cause poverty are intertwined, you should read this book about the eviction crisis in Milwaukee.”
- Believe Me: A Memoir of Love, Death, and Jazz Chickens, by Eddie Izzard. “Izzard’s personal story is fascinating: he survived a difficult childhood and worked relentlessly to overcome his lack of natural talent and become an international star.”
- The Sympathizer, by Viet Thanh Nguyen. “Most of the books I’ve read and movies I’ve seen about the Vietnam War focused on the American perspective. Nguyen’s award-winning novel offers much-needed insight into what it was like to be Vietnamese and caught between both sides.”
- Energy and Civilization: A History, by Vaclav Smil: ”He lays out how our need for energy has shaped human history from the era of donkey-powered mills to today’s quest for renewable energy.”
- The Best We Could Do, by Thi Bui. “This gorgeous graphic novel is a deeply personal memoir that explores what it means to be a parent and a refugee.”

12/15/2017

This is a light week:

Top Picks
- Advice For New and Junior Data Scientists, by a senior data scientist at AirBnB (and previously Twitter) https://medium.com/@rchang/advice-for-new-and-junior-data-scientists-2ab02396cf5b
  - Get on your mentor’s “critical path”, pick your tool based on the “language” of your problem, mechanisms for reinforcing your personal learning, find the right senior guidance, and take every opportunity to teach others.
  - Touches on “R vs. Python”. But no spoilers here ;)
AI: Advanced
- DeepMind Papers at NIPS 2017 https://deepmind.com/blog/deepmind-papers-nips-2017/
  - There are a lot of cool technical papers here, including new work on estimating the “uncertainty” in a neural net’s model, and deep reinforcement learning techniques for learning based off incremental human feedback.
Cryptos
- What the Hell Is an Initial Coin Offering? https://www.technologyreview.com/s/608799/what-the-hell-is-an-initial-coin-offering/
  - Good explanation of how new cryptos are “crowdfunded” in IPO-like events, and how the tokens can be used to represent more concepts than only currency
  - Note: total ICO funding has recently soared above $4 billion
- As it Turns Out, the SEC Doesn’t Completely Hate ICOs https://www.technologyreview.com/the-download/609756/as-it-turns-out-the-sec-doesnt-completely-hate-icos/
  - Further clarifies the SEC’s views: they will focus regulation on the class of new crypto tokens that are specifically meant to function like securities, but not necessarily on the more generic applications
Software
- Tail attacks on web applications: https://blog.acolyer.org/2017/12/11/tail-attacks-on-web-applications/
  - Introduces a new stealthy DDoS attack
  - “Tail attacks aim to create very short (hundreds of milliseconds) resource contention (e.g., CPU or disk I/O) with dependencies among distributed nodes, while giving an “unsaturated illusion” for state-of-the-art IDS/IPS tools leading to a higher level of stealthiness.”
- Apple Turi Create https://github.com/apple/turicreate
  - “You don't have to be a machine learning expert to add recommendations, object detection, image classification, image similarity or activity classification to your app.”
  - Designed to export models to Core ML for use in iOS, macOS, watchOS, and tvOS apps.
Business
- Disney Makes $52.4 Billion Deal for 21st Century Fox in Big Bet on Streaming https://www.nytimes.com/2017/12/14/business/dealbook/disney-fox-deal.html
  - Will undergo antitrust scrutiny

12/8/2017

This week: using AI to augment human intelligence, career advice that’s actually backed by evidence, and Numerai pays thousands of data scientists with cryptocurrency to give anonymized stock market predictions.

Top Picks
- 80,000 Hours: All the evidence-based advice we found on how to be successful in any job https://80000hours.org/career-guide/how-to-be-successful/
- Compilation of self-help advice in order of generality + strength of evidence
  - E.g. back-pain is one of your biggest threats to long-term productivity, so prioritize 1) excellent posture, 2) changing positions throughout the day, 3) exercise
- Using Artificial Intelligence to Augment Human Intelligence (credit to Matt Willian for finding) https://distill.pub/2017/aia/?source=techstories.org
  - Most people think of AI as offloading computation onto machines, but less time is spent thinking of ways to use AI to augment human work in rich ways with novel interfaces (this mirrors how our conception of computers has expanded significantly from “big calculator”)
  - Generative models (such as GANs or autoencoders) can be used to pull out higher-order abstractions in complex domains, opening the door to significantly more powerful interfaces
FinTech
- Numerai, a unique twist on crowd-sourced hedge funds (https://numer.ai):
  - Encrypted Data For Efficient Markets https://medium.com/numerai/encrypted-data-for-efficient-markets-fffbe9743ba8
    - A crowd-sourced hedge fund (with backing from Renaissance Tech’s founder) that distributes data with structure-preserving encryption (i.e. you have to build models without knowing what any of the time series represent!) and pays contributing data scientists with Ethereum-based tokens if their models make accurate predictions.
  - Super Intelligence for The Stock Market https://medium.com/numerai/invisible-super-intelligence-for-the-stock-market-3c64b57b244c
    - Combining hundreds of thousands of models from thousands of anonymous data scientists forms a powerful ensemble classifier capable of reducing error in stock market predictions
Software
- macOS High Sierra: Anyone can login as “root” with empty password https://news.ycombinator.com/item?id=15800676
  - Doesn’t look like this has been fixed yet; you can self-patch by adding a password to the root user manually
- Hindsight: understanding the evolution of UI vulnerabilities in mobile browsers https://blog.acolyer.org/2017/12/05/hindsight-understanding-the-evolution-of-ui-vulnerabilities-in-mobile-browsers/
  - The vast majority of mobile browsers are vulnerable to one or more of the attacks, 2) mobile browsers are on the whole getting less secure as time goes by, and 3) the popularity of a browser and its security are not necessarily correlated.
- Thing Many People Find Too Obvious to Have Told You Already https://news.ycombinator.com/item?id=15826445
  - Concise list of ideas about the software industry, such as: “Technical literacy in the broader population can be approximated with the Thanksgiving test: what sort of questions do you get at Thanksgiving?”
Business
- Stanford CS007: Personal Finance for Engineers https://cs007.blog/
  - Financial advice tailored to young software engineers, including common behavioral pitfalls
- Amazon unveils $250 AI camera and machine learning tools for businesses https://www.theverge.com/2017/11/29/16715688/amazon-ai-camera-developer-enterprise-tools
  - DeepLens is their new camera with AI on-device, while SageMaker brings end to end ML services to AWS - including not only TensorFlow/Caffe2Py/etc, but also off-the-shelf translation and transcription tools
  - More info: https://aws.amazon.com/sagemaker/
- Weather-tracker offers ray of sunshine for hedge funds: https://amp.ft.com/content/361457dc-2f38-11e6-a18d-a96ab29e3c95
  - From last summer: London-based Cumulus (now with $2.3bn AUM) has made 970% over the decade by finding arbitrage opportunities in predicting the weather
- 98,750,067,000,000 Reasons to Be Worried About 2018 https://www.bloomberg.com/gadfly/articles/2017-12-04/98-750-067-000-000-reasons-to-be-scared-about-2018
  - Like a mini-BDO of vague pessimism

12/1/2017

I’m piloting a weekly digest of interesting articles - let me know via email or the poll if you have feedback on format/content.

This week: contrasting perspectives on the future of AI, Quantopian has a public risk model, Coinbase receives both good and bad news, and Capsule Networks are poised to shake up image processing.

Top Picks
- Software 2.0, by Tesla’s AI Director: https://medium.com/@karpathy/software-2-0-a64152b37c35
  - Deep neural nets, now only in their infancy, will be the cornerstone of a new wave of software for problems where acquiring data is much cheaper than programming millions of variables.
  - DNNs have convenient properties (constant runtime/memory, easier to optimize, limited instruction set) and unintuitive shortcomings (hidden biases, common-sense failures). It may ultimately come down to choosing a diagnosable 90% accurate model, and a black box 99% accurate model.
- The impossibility of intelligence explosion, by François Chollet (Google-TensorFlow, author of Keras) https://medium.com/@francois.chollet/the-impossibility-of-intelligence-explosion-5be4a9eda6ec
  - “General” intelligence is ill-defined, as all intelligence is environment-specific
  - Recursively self-improving systems in the real world run into environmental constraints; most humans encounter productivity bottlenecks long before reaching 100% cognitive utilization
FinTech
- Quantopian Risk Model https://www.quantopian.com/risk-model
  - The crowd-sourced algorithmic trading hedge fund now has a risk model built into its online IDE/backtester
  - They also organize quant Meetups in NYC fairly regularly
  - Tutorial video https://youtu.be/JNEv_jHu6Is
Software
- Supercomputing poised for a massive speed boost https://www.nature.com/articles/d41586-017-07523-y
  - Exascale computers are on the horizon for the next few years, but it’s up to software engineers to write programs capable of harnessing this massive parallelisation
AI
- Concrete Problems in AI Safety: https://blog.acolyer.org/2017/11/29/concrete-problems-in-ai-safety/
  - Where AI can go wrong: Negative side effects, Reward hacking, Insufficient oversight, (Un)safe exploration, and Fragility in the face of distributional shift
AI: Advanced
- Dynamic routing between capsules: https://blog.acolyer.org/2017/11/13/dynamic-routing-between-capsules/
  - The new neural net architecture of Capsule Networks (CapsNets) has been causing a lot of excitement for its potential to displace Convolutional Neural Networks as the preferred method for image-related work, due to CapsNets ability to handle arbitrary pose transformations
Crypto
- Bitcoin exchange Coinbase has more users than stock brokerage Schwab: https://www.cnbc.com/2017/11/27/bitcoin-exchange-coinbase-has-more-users-than-stock-brokerage-schwab.html
- Coinbase ordered to report 14,355 users to the IRS https://www.theverge.com/2017/11/29/16717416/us-coinbase-irs-records
  - Anyone moving more than $20,000 on the platform is subject to the new order
  - This is because only a few hundred users have been reporting gains on their taxes
Business
- Stop Using Excel, Finance Chiefs Tell Staffs https://www.wsj.com/articles/stop-using-excel-finance-chiefs-tell-staffs-1511346601
  - A growing number of CFOs are having trouble collaborating efficiently using Excel, and are migrating to cloud-based solutions
- Personal Sedan Sales in Jeopardy as U.S. Auto Market Transitions to “Islands” of Autonomous Mobility: https://home.kpmg.com/us/en/home/media/press-releases/2017/11/personal-sedan-sales-in-jeopardy-as-us-auto-market-transitions-to-islands-of-autonomous-mobility-kpmg-research.html
  - “...adoption of this new transportation mode will not be immediate, and it will not be everywhere. Instead, it will arrive metro market by metro market in what we call ‘islands of autonomy.’ Each island will need a unique mix of vehicles to meet unique demands”
- Uber Paid Hackers to Delete Stolen Data on 57 Million People https://www.bloomberg.com/news/articles/2017-11-21/uber-concealed-cyberattack-that-exposed-57-million-people-s-data
  - Uber has already had a tough year, and now they have to provide free credit monitoring to millions of drivers
- What I Learned From Reading Every Amazon Shareholders Letter: https://blog.usejournal.com/what-i-learned-from-reading-every-amazon-shareholders-letter-cdc35f309e8b
  - Interesting takeaways: Amazon’s own set of “Principles”, the focus on maximizing free cash flow, staying efficient as the workforce grows, avoiding the Innovator’s Dilemma, taking asymmetric risks, and executing.
Fun
- I turned a Furby into an Amazon Echo. Introducing: Furlexa https://howchoo.com/g/otewzwmwnzb/amazon-echo-furby-using-raspberry-pi-furlexa
  - Should only take $50 and 2-3 hours to do it yourself
- HQ Trivia: https://techcrunch.com/2017/10/17/hq-trivia/
  - From the founders of Vine: A daily live trivia game with cash prizes