evilturnip

This is interesting. He seems to imply that the AI will just provide the solution. He's concerned that the process of getting to that solution is the important bit. But I can see two versions of "process".

A.) The actual steps of the proof, which I assume the AI would provide. B.) People, while working toward a solution, finding novel properties/methods along the way that open new avenues of research + new open problems.

Does B.) actually happen? Would knowing the solution to a problem stymie the process of finding new open questions? I would assume finding a solution may unlock other problems too. So maybe on balance it's not really bad?

I guess in the end, I'm just making the obvious case "the future is uncertain in the face of AI".

Can you use postgres as a state store for a distributed application?

It seems this article is trending toward that view: If you can maintain transactional consistency along with application workflow state, then would this generalize to maintaining distributed application state in general?

The follow-up would be: Would this be preferable to Valkey/Redis?

Most people don't have an audience they would share it to if we're being honest.

If there's a article/site I'd be interested in sharing, it might be to a slack channel or a text message, in which case I just copy/paste the URL.

I do feel like for all their dogfooding of AI coding, their own software/APIs are quite buggy and work against their message.

Claude Code is especially buggy in windows terminal. The rendering is quite slow, choppy and lines frequently get garbled.

In contrast, using antigravity cli is the exact opposite: fast, smooth and very responsive.

I suspect being a "better programmer" cannot be said unequivocally at their level. At that percentile of achievement, it depends on the specific dimension you are talking about. It's true of the highest skill in any field.

It's obvious that those that write the tools/infrastructure are less visible than those that create the end product.

I don't know a single name behind the construction of the AI tensor core in Nvidia's chips but it is effectively what runs all of AI.

I work on the open-sourced Jedi Academy game engine, which is based on id tech 3. It has special logic to specifically adjust the gamma of the physical monitor during launch, then restore it upon exit.

This is why when these older games crashed on your PC, the monitor would look all washed out due to the manual gamma adjustments the game made that didn't get restored.

Ok, stupid question, but what applications is something like tailscale/iroh used for? I've never worked with this type of tech so curious where it is valuable.

Masters of Doom is a great book on the history of id software, which includes the origins of the development of smooth scrolling by Carmack and Romero, which was groundbreaking at the time on PC.

I feel like the very thing that EU is great at: more consumer friendly and anti-corporate legistation, also hamstrings it for innovation. Why would VCs invest in AI there instead of in the US where they don't have to worry about any of that.

Ray tracing is one of those problems that is conceptually so simple, yet continues to take so much mindshare because of all the challenges to implementation.

"Think of the children" used to shoehorn in the police state. Shudder to think where this will be in 100 years.

Yes, growing at average 15% a month will make you a billionaire in so many months. But somewhere along that trajectory you've tipped the balance from "eveyone is benefitting" to "in order to grow larger we need to exploit and extract."

Clearly nothing is universally the case, but this pattern repeats in enough freqeuncy that it's effectively the case.

On the other hand, the fact that exif is such a widespread standard has been a great tool for photo archaeology.

People who can track down the original exif can recreate when, where and with what equipment the photo was taken. It's been great to identify places and people for posterity.

I do find the BDFL approach much better for language design. You might disagree with the direction of the language, but there is usually a "philosophy" or "taste" driven by one person that tends to be consistent over time.

In fact, I think Guido himself resigned due to the experience he had trying to get a PEP through the committee.

It's great that we are getting so many open source model releases, but I just feel like SOTA models will always be in the hands of the big players. The hardware requirement to achieve SOTA are just too steep.

My alternate universe would involve some sort of decentralized investing scheme to build data centers running massive open source models that could compete on some level with Anthropic, OpenAI, etc.

Arabic script is a great test to see if your terminal/renderer/UI can handle anything: contexual shaping, cursive connectivity, bidirectional text layout, diacritics and vertical displacement.

I went down this rabbit-hole awhile back and it made me really appreciate the complexity of the script.

I think it’s impressive that an LLM can take you to a local maxima in one-shot.

But once you start maintaining it, improving it and fixing bugs, you’ll eventually need to rip it apart and put it back together again while understanding how it all works.

This is why I think the better approach isn’t to one-shot but to have the architecture in your head and build it up piece by piece, with the AI accelerating the code writing.

Of course the tired follow-up: “But if the government was functioning properly it would only do the things I want”.

This whole thing is comedy.

Anthropic pretending Mythos 5 is so capable it's going to destroy everything, but will release it anyway with "safeguards" (when does this ever work?).

US Gov't using this fake hype as an excuse to handicap Anthropic simply because they have a vendetta.

I always wondered: If Fable 5 uses less tokens, would that be enough to offset the increased compute cost? The reasoning for pulling Fable is that they wouldn't be able to handle the compute demand.

Even if not, I wouldn't mind waiting longer for responses when load increases, knowing Opus is always around in case I need a faster alternative.

"Having an agent spit out generic software architecture is probably better than what they were producing before."

If they were a poor programmer/architect, I don't think the AI would make the end result any better. It would amplify their lack of skill. Sure, the low-level code might be more airtight and idiomatic, but that's not even where poor skill really manifests itself. It's at the higher level of thinking in terms of the system and understanding the proper context of the business/technology, etc.

This is more or less what I do. Then again, I work on a small parts of the codebase at a time, so maybe the autonomous agent works better when you're doing larger refactors over large codebases.

Even in that situation, I think I would still only feel comfortable approaching the task as I would do it without AI, and using the AI to accelerate the parts that would be time-consuming. E.g. finding where/how feature X is implemented, how it would affect the overall system if I were to change it this way, etc.

Working with AI is trying to reduce the probability it'll pick undesirable paths. It's an exercise in trying to avoid what you DON'T want.

I suppose it's the same as asking someone else to take care of a feature and hoping they understand what you have in mind. The difference is that there's a lot of context that's shared between you and a human developer that is simply absent with AI.

I get exactly what you mean. After getting frustrated with how slow Claude was on my personal projects, I switched to Google Antigravity with Flash models and the speed difference is huge. I feel more in the flow and just more focused on the task. I did not realize how much a difference speed can make.

Claude is better for extremely complicated, large codebases where its slower response time might be a good trade-off for the complexity of the task. Antigravity and other fast models works so much better for smaller projects where you want a "flowy" code, run, debug cycle.

I also ask it to explain the system back to me. Obviously it should understand the system just by reading the code. But somehow, explaining the system back to me seems to make it more effective. Then I'll ask it questions about how I should make changes to the system. Sometimes I'll agree, sometimes I'll disagree and offer an alternative and ask it to assess the alternative. Having this entire conversion in its context seems to make it way more effective at refactoring/unsloppifying code.

Probably thinks you were talking about two-stage ICBMs.

Vulkan is horrendous, LLMs largely eased the frustration of working with it for me. We talk about AI slop, but what about the human slop...

nowgrep is supposedly even faster than ripgrep:

https://x.com/CharlieMQV/status/1972647630653227054