StilesCrisis

The conclusion is simply wrong. `union` should not be used for type-punning in C++, or in C code which might be compiled by a C++ compiler.

I can't downvote posts yet, but if you have the ability, consider it. This is clearly LLM slop without human review.

I've seen LLMs "connect the dots" across complex systems many times before. When it works, it's shocking how quickly it can pin down a bug that spans across the software stack.

1M context window is plenty. Once it's skimmed the code and come up with a theory for the problem, it can spin up a subagent that has a whole fresh context window and it can dedicate the whole thing to that one hunch.

Someone did it!

https://youtu.be/7NfyZhV1dKM?is=YOUXCHuFUiPdlD0p

Read to the end: they aren't actually looking for the best-compressing output, because this quickly devolves into aaaaaaaaaaa. They keep a sliding window over a small portion of recent text and use that.

Basically I think the entire premise falls apart due to that choice--they forced an interesting-looking outcome by adjusting the algorithm until gzip started picking random slabs of letters instead of ever-larger repeating runs.

Shorts aren't all bad. I feel like they've slowly drifted to regular-video length, and the major difference is just the tall aspect ratio.

I am on the exact same page as you, but there is definitely a market for LLMs which speak more conversationally and less like Claude! Non-programming use cases abound and most users don't like the rigid, exact tone that engineering demands.

Makes sense. When Google Jules was new, I experimented with it for a while and it was embarrassing how often it flubbed tool calls, often repeatedly. It tended to forget the working directory so almost every command needed two tries.

A few weeks ago two "Claude wrapper" projects showed up here within 24 hours of each other. Both seemed to just sand the details away--useful if Claude is too wordy, I suppose, but seemed to also lose real substance.

When I look at Claude's memory markdown it tends to list dates for significant requests already, so it can untangle this sort of thing. And if it can't it will just ask directly.

> Do you want to watch a video? YouTube knows you like cooking and art streams. But why not also recommend a few clips about the stock market bubble, global warming, and the war in Iran.

I'm always fascinated when people claim YouTube pushes content like this. It's an extremely common take, but I don't think it's true.

My experience has been that if I stop clicking something, the algorithm soon stops recommending it. It occasionally brings back a blast from the past--e.g. I watched a sumo competition once, and once in a blue moon it recommends another. But you can downvote bad recommendations, and I simply don't get recommendations for ideas I've never engaged with before.

FWIW I'm not saying the algorithm is great. My feed is absolutely worthless right now, but it's all goofy Japanese content because one of my kids is extremely into that right now. There's zero political content because she isn't showing any interest in it. Heck, most of the stuff isn't even in English anymore.

If you can make an agent reliably follow instructions, you've solved alignment completely. So I'd be particularly surprised if you can square that circle as a hobby project!

For sure, actually mirroring the shape of inlined code properly rather than just emitting code N times is already extra credit as far as I'm concerned. I've got enough loose ends without chasing that sort of thing!

I don't think that's right. I helped work on SNES9X in the early 2000s and it ran nearly everything perfectly.

Sure there were smaller emulators which didn't work well, that's understandable. But gamers didn't need more than one or two emulators to play every game that existed; you just needed to pick the right ones.

Claude can definitely wait 50 seconds and then do a second thing. I see it all the time when running tools that never stop but iterate towards a goal. It can either run a bash `sleep 50` or use the ScheduleWakeup tool.

Is the date really useful? Usually I want the AI to remember rules, like "always do X before committing Y" and a timestamp doesn't change anything there.

Float constants are stored as data which is deduplicated on a per-TU basis. So if my TU has 1.0 and 0.0 in it, every function in that TU which uses those numbers will load from the same address. The next TU will get its own dedicated 1.0 and 0.0 constants.

We're not ten years early. If you can afford the tokens, and can course-correct the agent when it gets off in the weeds, AI can do 95% of the grunt work of a decomp now.

Also, "widescreen OOT" is probably tweaking a pair of projection matrix floats in the binary so that might not be a great example :)

I don't see why the goal of a decomp needs to be "learning how to decomp"--although I guarantee you that working on one, even AI-assisted, will teach you those skills.

Personally, I think it's fascinating to see how the industry greats actually made the magic happen. Everyone talks about the endless fine-tuning of controls that Miyamoto asked for in SM64: now we get to see what that actually looks like! Or maybe the game had a hidden secret debug menu that no one's ever seen in 20 years, or an unfinished character is hiding in the code.

Maybe my perspective is different because I've worked on games for most of my career, so seeing how _other_ teams solved the hard challenges is exciting for me.

Yup. Unfortunately the project I'm working on does not have any debug info. There are not even lingering __FILE__ strings anywhere. Just picking out the TU bounds has been a chore!

It's a huge benefit that things like the OS and SDK are known and predictable. This got me a foothold. Otherwise I would have gotten nowhere.

I'm still impressed that this decomp has macros (fully lost in the assembly) and nearly-100% meaningful variable names and struct layouts. That's not easy!

Pretty sure the Kremlin just murders folks that they have a problem with.

Working on a decomp hobby project now, and "100% matching" on an individual function can be so deceptive! These two functions could look identical in assembly:

void foo() { bar(); }

int foo(int x, float y) { return bar(x, y); }

This is because parameter-passing and return-value handling can sometimes be done in zero instructions in PPC, if the values are already in the right registers. So you need to continually go back and reassess old functions once you learn the signatures of newer ones!

You're right. I said "the accountants didn't ALL disappear" but surely, some of them had to endure a career change. But what's the alternative?

As much as it would suck to be one of the 27 unemployed, I don't think the answer is to throw away our calculators and start doing our bookkeeping by hand again. I don't think you need to feel too bad about this. The accountants didn't all disappear when Excel showed up; they just got to work on the more interesting parts of the job, instead of rectifying arithmetic.

You would also need to do it sixty times a second. Games already try to do the "easy" culling of hidden geo, and you don't automatically know the screen-space bounds of the things you are drawing, like you generally would in Skia.

Game development has apocryphal stories aplenty about inefficient designs--e.g. the thousand-polygon model of a screw used on every rivet for every crate, or making a ship-in-a-bottle by taking a full sized pirate ship model and scaling it down to 1% size. Usually the solution is just to stop doing that.

You can just look at the commit stream and see for yourself. Despite community pearl-clutching, 99% of the work is just good old fashioned software engineering.

Hard forking doesn't solve the problem of closed source.

If the goal of a hard fork is "the community will invest equal resources that Google does today" then you wildly underestimate how many engineers work on Chrome.

Americans don't use Celsius for anything except scientists talking with other scientists.

Not at all! Cars are an extremely common trigger for anxiety. You are far from alone in this.

Or rather, the non-failure case!