StilesCrisis

"If someone showed you that heatmap without the game labels, I think you’d be able to immediately identify at least the first three games" -- absolutely not. Pac-Man in particular seems like a huge stretch to me.

It really depends, right? For some code bases, the author would be pissed if the agent started ripping up all the existing classes. This is where judgment comes in and why you and I are still employable.

> the worst thing you can do in QA is miss a real bug and ship broken code to prod

Maybe it varies by domain but I'm quite used to codebases where "fix every discovered bug" is basically infeasible; finding 1000 miniscule bugs isn't helpful. Finding _serious_ bugs is the goal.

All the "competition" is just distilling the frontier, so it makes sense that they are perpetually behind by months.

Hypnosis and optical illusions are adversarial attacks on shortcuts our brain takes.

Qwen in reasoning mode is ridiculously anxious and ends up going in circles. It's not just math, it's how it behaves in general.

Others have also fine-tuned Qwen to shake this behavior already, so it's honestly not entirely shocking that Astra could too, but I think this is the first time it was done "by hand" and not via fine-tune.

If you look at the kind of bugs coming in, it's from all corners of the codebase, not just things changed in the last 1-2 years.

If you are at the RC stage and still uncovering segfaults, "rip the feature out" is probably the only choice.

You think LLMs don't make mistakes? My sweet summer child! I want all the guardrails with LLM code. Reviews are fatiguing. If I can remove "memory safe" from my list of worries, so much the better.

Unfortunately, I don't think history actually substantiates the claim that users evaluate all their options and migrate away from systems as they enshittify.

Can you just keep fine-tuning indefinitely? I don't really know but it feels like eventually that might start impinging on the model's abilities?

> You don't need bigger or better models. Current models are more than adequate for any task.

At the very least we need _current_ models, though. The weights have all sorts of information baked into them, like "how do Python unit test drivers work" or "what does idiomatic C# look like," that will just naturally go stale over time as the frameworks improve. In five years, Qwen 3.8 will still be targeting the Python of 2026.

If Qwen is so great, why did they use Claude to do the write-up? This text is pure, unfiltered Claude. (e.g. "Ten problems, five passes, not the authors' 100. On a sample this small, 9.8 and 9.6 are the same number. The claim is parity with their harness, not a gain on it.")

Zerg rush lets you build fast. "I can build anything at 10x speed" doesn't mean you can actually change user behavior. That's not an engineering problem, it's mostly a marketing problem, and you probably can't out-market Google. They own the ad platform!

They moved it to HIToolbox when Carbon framework was eliminated.

This was never provided by classic Mac OS and many apps hacked it in by hooking Get/WaitNextEvent. In Carbon they didn't support this sort of "mangle the global system callbacks" shenanigans, but there was a real developer expectation of this API, and Apple was begging developers to port to Carbon. This API filled that gap.

You may have heard this from Culture War-adjacent content. In reality, this is wrong, and you would struggle to find a real employed academic who agrees with you.

Also, languages aren't dictated by a standards body like C++ is. The rules are written to capture the way the language is actually spoken. If the rules diverge from the language, the rules are actually the thing that is incorrect! This is the difference between "de jure" and "de facto" rules.

RegisterEventHotKey didn't actually exist in OS 9. It was added in CarbonLib only.

Anyone could build a search engine too, and many did, but none could dethrone Google in the end.

Interesting! I should investigate that. I was following their docs and GitHub was 100% the happy path, and I was actually hoping to avoid it (downtime woes).

I mean for sure, this isn't an OS that _stores data_. It'll pull up any data you ask for, though!

VibeOS? It's been done! It's hilarious: https://youtu.be/7NfyZhV1dKM?is=Y8LwTbY_stLPilqR

I wanted to set up a Cloudflare static page and all the tooling is completely designed around Github. If you want any other service, you have to roll it yourself.

Does it? I am pretty sure the entire Carbon stack is retired now.

Agner Fog is also discussing CPU architecture, not high-level language. The whole point is that the optimizer _already_ takes knows these things and rewrites your code automatically. If you write a zero-to-100 loop in C, that doesn’t mean the assembly will literally show a value going from zero to 100; it may unroll, or reverse the loop order, or any weird combination of things that the optimizer has decided will be equivalent and fast.

Optimizers eat stuff like this up. It'll restructure your loop however it wants.

It's just spacing. X-- > 0 and X --> 0 are the same thing in C.

Do LLMs really care?

"They" has been a gender-neutral singular or plural pronoun for hundreds of years.

Makes sense though! It's really tough to test touch control programmatically, and that's all LLMs have. Same reason ChatGPT can't play Super Mario Bros.

When Knuth wrote that, I don't think dependencies as we think of them today existed. Libraries were few and solved specialized, difficult problems.