mpoteat

These are pretty surprising claims... As far as I know, V8 itself is written in C++, is this self-hosting parser faster than Node's parser?

Volley (YC W18) | San Francisco, CA | INTERNS | ONSITE | Typescript

The Volley Infrastructure team works on internal tooling, processes, and engines that power Volley's voice-powered games and media applications. The work you'll do will be high-impact, facilitating our entire organization to achieve its mission objective: ship voice-applications quickly and easily.

Practically speaking, the work will involve plenty of backend systems engineering, mostly in Typescript, with a strong functional bent, on AWS.

Entry-level NodeJS developers are encouraged to apply, especially those with interest in Haskell, Idris, or other type-heavy languages.

Mention Hacker News when you apply.

https://www.ycombinator.com/secret-companies/1826

I would maybe recommend Typescript if you want to harness the type level power of "proving" correctness for certain aspects of your code.

It's kind of a shame that this subset of research is kind of stuck on C as a standard. I feel like higher abstraction languages would potentially be easier to reason about, as well as personally more useful to me.

Sure I know C11 but it's just not a tool that solves the problems I'm interested in.

Every developer should do all of these things, the behavior and norms described in this blog post seem like the absolute minimum to have a healthy work environment.

It might be more expensive in the long run, but it's an operational expense rather than a capital expense, which I would argue most software projects are.

If you like terse code I recommend APL.

Probably because it will incentive more people to catch lobster.

Search for Fabrice Bellard.

If so many services use Let's Encrypt, do you think there's a risk they become a target for sneaky surveillance activities?

The million dollar question is if it's possible to construct a theory of computation where the input language itself is automatically differentiable, and where the execution semantics are also so. Perhaps a continuous spatial automata that has been proven to be Turing complete.

I view the usefulness as expanding the class of programmatic functions we can differentiate. One usefulness of differentiating programmatic functions is the ability to perform gradient descent optimization on that function... which has applications in operations research for example.

The crossed out parts of failed revolutions of tech migration was brutal to read. It really underlined the article's message.

GPT2 does have a goal in mind, namely an idea akin to "minimal surprise". The problem is that it doesn't have a model of the world its talking about, so it quickly says things that are locally sensible but globally inconsistent.

There's evidence that signal differences in neutrons in the visual cortex is used for object recognition, whose behavior gives rise to some of the optical illusions we experience. Weak evidence for temporal differentiation also having the same underlying mechanism.

After skimming through this, it seems fair to say a lot of the research to come out of this group has been relatively "soft", i.e. ethics and interpretability rather than e.g. the newest most powerful models or innovations like gauge CNNs.

Not bad, just surprising.

Intel is a big company, but this also has the background context of their fierce competition (which arguably they are losing) with other chip makers in the enthusiast space. Maybe would explain an attempt for more R&D green field projects.

Not a full time ML researcher, but I thought I understood that batching is already an extremely common practice. I don't see the novelty here.

Yep. Puppeteer is the state of the art wrt headless browsers. There's money to be made selling B2B wrappers of OS software though. Not to trivialize the author's work...

The average HN user might balk on an early sign up wall (and become very upset), but that doesn't preclude it being the optimal placement for most undifferentiated users of your metric is purely the number of sign ups completed.

This app / clinical study reminds me strongly of an old Flash application "Click the color and... Not the word". So there is some prior art here - although this application has much better packaging.

Success and life satisfaction (shouldn't be) zero sum.

Personally, I don't expect FDIC insurance on swap space used for trading, or generally anything except my bank account.

Github can host pdf.

ACH is a bank to bank transfer.

This is really very cool, and makes me want to pick up Coq again. I know the set of formally verified proofs is very small as compared to traditional mathematics, but I think Coq and its cousins are vital for a future revolution in mathematics.

As it turns out, just because a window is left unlocked doesn't mean you can just burglarize a place.

This is QWERTY, albeit not one linear rectangle.

Plus, also incentives research.

Code golf is anecdotal, at least, evidence that proving lower bounds on the Kolgomorov complexity of a particular datum requires intelligence.