Hacker Newsnew | past | comments | ask | show | jobs | submit | rfgplk's commentslogin

OpenAI's models will actually obey a "be concise, leave no comments rule".

Yeah, still abysmal. If you have the time or tokens, see if it will obey an explicit "LEAVE NO COMMENTS WHATSOEVER" command. Opus 5/Fable 5 outright ignored it.

Opus is only usable if you have a post-turn formatter that strips all comments from the generated source. I'm not even kidding it's that bad.

So still effectively nonsense.

> Rewrites are declared by the publisher, never inferred from overlap.

This style of writing is idiotic because it conveys no additional information. It's no different from stating

> Rewrites are declared by the publisher, never when moons collide.

The two sentences are actually logically identical. No idea why these models keep writing like this.

> Folds are monotone single statements. gen only goes up, extents only grow, processed periods only union, rw_start only moves earlier.

This is even more ridiculous.


0% chance, without LLMs. WASM-capable engines would require years if not decades to build, even with expert level knowledge. See Jonathan Blow who has been working on his engine for a decade now, and he's arguably of the most talented engineers who ever lived.

Jonathan Blow is writing his own language as well, and is also already successful enough that his work is just a hobby he can take at any pace with no urgency. A WASM engine is really, really not that difficult. At its core, you need rendering to a canvas, audio, keyboard/mouse/gamepad handling, asset loading/file saving, and an update loop. Writing this code is mostly not different from writing code in other languages, since you are literally writing other languages that happen to compile to WASM, the only differences being that you need your one-time WASM toolchain setup, some JS glue interop (which is not really different from needing C interop for native engines written in languages other than C), and to be aware of browser pecularities regarding file access, update loop, threads, etc. which you also have to deal with if you ever wrote a JS game anyways. After you have those core elements in place, everything above that is isolated in a game/engine logic abstraction layer that isn't any different from writing native code.

If you want to place a bet on it, we can do a $10,000 bet in escrow contingent on myself implementing a well-specified WASM engine from scratch on stream without LLM usage in a month. I would love an opportunity to demonstrate how wrong you are. That said, rather than taking your money, I could also just share a streamer's content with you[1]. He implemented 3D web rendering with no dependencies in a 20 minute lecture, and it would take 10 minutes if you were seriously focused on doing it quickly. Sure, it was rudimentary pure JS rather than WASM, but really consider whether you think this 10 minute exercise couldn't be done in a language that compiles to WASM with 160 hours, while including the other hardware/OS-layer abstractions aforementioned. On the other hand, please do take me up on my offer. You said it's a 0% chance, after all -- surely you don't want to pass up on the easiest $10,000 of your life...?

[1] https://www.youtube.com/watch?v=qjWkNZ0SXfo


__builtin_memcpy is constexpr under clang with constraints

> (Feels a bit like LLM junk, but honestly not sure.)

Why leave such disparaging comments?


Because if the comment is correct, it's useful signal for other potential readers to know that they can skip it.

I try to be as charitable as possible when reading and commenting online, but we don't have infinite attention and thanks to AI, it's easier than ever to end up wasting it on things that aren't worth reading.

(I'm not claiming that the article here is or is not an example of that.)


Even if it's not from an LLM, it's still wrong! You've focused on the last important of that comment, really.

Correct, but they should explicitly announce this ahead of time.

A general rule of corporate behavior unless they are forced to under duress.

If this is duress of competition or at gunpoint of regulators is up for the population to decide.


Luckily for us open weight models exist. Until regulatory capture anyway.

Unfortunately there's just no demand for that. People have suspected they do this for years, but they get shouted down for even suggesting it.

I find it odd, because it's not just plausible, it's understandable that they would need to dynamically throttle these things or degrade people to a different model. I would be fine with that if there was transparency. I'd think a business building any type of logic on AI decision making would like to know if they are actually using Opus and not Fable. But apparently the idea that these companies don't have infinite resources is conspiratorial fear mongering to most AI boosters.


> I have no hard data but I have a strong feeling this morning that something's wrong with Fable 5 compared to Friday evening.

Fable is effectively worse than Opus 4.6 now. They severely messed with the model.


Yep, that's what they've been doing for a long while now. Also the amount of tokens you get per sub varies drastically from month to month. Needs to be regulated.

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: