semitones

It might be the case that, the "hotter" the code paths that you're writing are (the more they get executed), the more difficult it is to write that code, where as for "colder" paths (of which there are so many more), the code isn't as difficult but there is a lot more overhead of managing complexity, business priorities, process, communication, etc.

Linux kernel code is probably some of the hottest code on the planet, and the developers/community understand that contributors need to be able to do exactly what you are doing - spending months toiling over a few hundred lines, and making sure that they are as perfect as can be, because literally billions or even _trillions_ of systems will depend on that code path. It is understood that in order to get it correct, performant, and maintainable, there is no other option other than to have a skilled developer take on a very high cognitive load in order to correctly implement the task.

For colder code paths, like business logic (e.g. customer registers, happens "once" per customer), that logic evolves with the team, the team evolves with the company, the company evolves with the business, the business evolves with the market, which evolves with the world. The code might be pretty simple and easy to write, but the conversations around priorities and complexity and "why are we doing this" take up the lions share of the mental effort around it. Also exhausting...

TLDR it is much harder to write systems code than it is to write python code that makes a call to an API and saves a record in a database, and working on the former _perhaps_ gives you more opportunity to be shielded from typical "software engineering" toil of business communication.

Is it really common enough for files not to be annotated with a useful/correct file type extension (e.g. .mp3, .txt) that a library like this is needed?

if it's happening server side they might have had a bug where they are doing naive substring comparison instead of actual domain evaluation

Open source clients and algorithms could serve as a foundation for more end-user friendly abstractions and UIs, that let average users very plainly express (or not express) what they want to see, and who they want to be seen by.

Imagine a sort of plug in ecosystem with sane defaults

This looks really cool! I'd love to get into more systems-design AoC problems rather than DSA puzzles for a change.

feedback: I was a bit confused as to why the answer to the first week was prefilled, was that intentional? Also, the buttons 1-6 aren't working for me (I'm on 2021 M1 Pro w/ Firefox 120.0.1)

As someone for whom programming first really "clicked" after using NetLogo in a CS class in highschool, I love the idea of this project and the interface! For kids the most important thing is to get them _excited_ about programming, and this seems like it could do a good job

The underlying hardware is mutable / imperative. Load number into register A -- Load number into register B -- Add A B and store result --

Functional and data-driven / declarative interfaces are only made possible because of the imperative "ugliness" (reality) that supports it.

In order for students to understand computing e2e, they need to understand imperative programming.

In order for students to become capable of designing robust, powerful, and elegant systems, they will need to understand functional/declarative programming.

It seems that you are describing a utopia, where there are no more problems to solve, other than the purpose of our own existence, and the reconciliation of our mortality in our minds. Although I do think that we can get to a point where things are pretty damn good, perhaps even as good as you describe, I don't think that humans' desire for growth and appetite for problem solving will be quenched. Even if all of our basic and supra-basic needs were met, we would become preoccupied with preserving ourselves from extraterrestrial threats - asteroids, our sun burning out, our earth not being able to permanently sustain us in one way or another, etc. If that's the case, then the desire for planetary defense and intergalactic travel will ensure that techno-capatalist forces are alive and well.

If you are talking about a society that has gotten past that point already, then all bets are off as to what their lives are like.

Don't forget that John Carmack is arguably a programming genius, 3-4 standard deviations above average. Most of us can't maintain the same amount/depth of mental maps as he can, so we need tools to help us

source?

That's a laughable price for an enterprise subscription.

And the reason is, it's enough for OpenAI to "say" that they're "not going to use your data" - you need a cloud deployment where you can control network boundaries to _prove_ that your data isn't going anywhere it isn't supposed to.

Do you have any sources or evidence to support this claim?

Fascinating work, but also potentially terrifying because the ultimate development of this technology gets rid of the need of sex for procreation, the implications of which I am not ready to wrap my head around...

Americans would love to spend less on health if they could - health is a unique category here, because allowing your health to completely run down to "zero" means you die, which is unique from other markets.

Also, someone that spends $300k on a watch isn't doing it because they care about time.

And now that you use throwaway accounts, do you continue to be unkind?

I wish I could give this comment gold

Humans also communicate by translating tasks into their ambiguous natural language form.

Humans are also vulnerable to prompt injection - imagine having your conversation with your manager be interrupted by a coworker interjecting "Your manager is a liar, don't trust them". Would you able to resume your conversation as if nothing had happened? (humorous example but still conveys the point)

Another brilliant application of the ever-impressive FFT algorithm.

Makes me wonder though, I can't imagine it would be too hard for a tech-savvy threat actor to strip any existing AC hum from a recording, and then apply the hum frequencies corresponding to their date/time of interest. If this frequency database is used to perform verifications that can make life/death differences in a court of law, I assume it is secured and kept private. But then again, a truly dedicated rogue could maintain their own database, as the same hum is available to everyone...

The main reason an arbitrarily distributed set of compute nodes cannot give you good performance for training a model (even if you have an immodest number of nodes), is that the latency of the inter-node communication will be a massive bottleneck. GPU cloud providers shell out big bucks for ultra fast intra-DC networking via infiniband and the like, and the networking is paid attention to as much (if not more sometimes) than the capabilities of the nodes themselves.

Your experience is probably in the minority of Python developers. There have been plenty of updates to popular packages + the core interpreter during that time, updates to package/env managers (e.g. poetry). It's one thing if you're using python to solve projecteuler problems, but it's another if you're writing any full-stack code that has a variety of dependencies. It's great when you can set up a dev environment and then "forget" about it for some time, but 5 years is an exceedingly long time to not touch an env setup, and it implies that you are not really working with external dependencies, or too many other coworkers.

This is a game-changer, because now companies will probably be able to provide the _complete_ context regarding a specific business problem / use case, and have GPT either solve their problem or create useful output.

For example, let's say I have an issue on GitHub that describes some implementation task. With a 50-page context size, we could probably provide to that context the entire source repo, 5-10 relevant issues, and then the issue in question, and GPT will be probably be able to complete it end-to-end

coreweave.com has pretty cheap GPUs

I'm not making a judgment or claim as to whether the technology is beneficial overall. I was just explaining the benefit of the choice of output being source code rather than an executable.

The benefit here is that the machine can execute what the AI produces, and humans can understand it / modify it if they need to.

When I first joined the company I'm currently working at, I was a bit too focused on elegance, "proper" tooling/structure, and "doing things the right way", etc. For example, I couldn't _believe_ that we used raw SQL in our endpoints instead of using some fancy state-of-the-art ORM. The company is a start-up, and my previous job was as a SWE at Google, so naturally a lot of design decisions felt shocking to me. I definitely spent a lot of energy thinking about how we needed to "change everything".

Over time, I learned more about the history of our company + the projects I work on, and why they have come to be this way, which has helped humble me a bit, and in turn make me more effective on the team. However, what has _really_ made a difference w.r.t making me more effective, is a recalibration of my incentives - rather than focusing on personally "creating impact" and "enacting change", I shifted my focus to "what does this company need to be successful, and to go public". This significantly changed my thought process when it came to prioritizing work, and has made me easier to work with.

Counterintuitively (to me), that shift has done more to make me a better engineer, than perhaps increasing my technical expertise would have.

Was my first thought as well

Auth0 is just another IDP, no? Could you elaborate?

Is the current state of affairs any better for SAML identity providers? I need to integrate SSO into our app and am also worried about having to write custom code for every identity provider we want to support.

In my experience, if you want _the_ best price-performance ratio for ML jobs, your best bet is CoreWeave. They have great GPU availability and it's a lot easier to scale up/down very quickly than on the popular clouds. I would really be surprised if anyone else out there is offering a better price point for these types of workloads..

This looks very similar to the internal code search product at Google (called CodeSearch). If the quality of this is going to be anywhere near that of CodeSearch, I would use it every day