vatsachak

I said gcc, not a C compiler

Okay tell me why that's not even close to gcc. You can use an LLM

Math academia 2025

> Sorry, only epic problem solvers allowed here

Math academia 2026

> We were more than just problem solvers

I think people are overblowing this though. Wake me up when GPT-whatever writes gcc from scratch, then by the Curry-Howard I'd be impressed

Timothy Gower

Gr8 b8 m8

Muh academic ivory tower

That's awesome! Congratulations!

I'd imagine that in three months when we all have access to communicating agent swarms this should be easier

After AI agents get good enough the real bottleneck will be power generation and political systems.

I think the idea would be making a frontier model that does this. One that is trained on multiple queries

I agree with the caveat that it's more like a cracked junior engineer who can manage swarms of interns.

Frontier Labs will probably survive off hype valuations but will serve the important purpose of discovering architectures/techniques that will probably spread through rumors/transfers to the rest of the world.

Yeah it would be quite trivial to try and implement an auto regressive AST generator for STLC with Jev provided that you had bounded variable names and integers.

As you said, if it worked, they would have demoed it haha

If you provide it an AST of the english language, yes.

It's not an LLM though it's a frontier model on structured data

Cool project!

That's what I was insinuating through "better encoder"; the model creating more efficient representations of ASTs using something like JEPA

It could be used for coding if you gave it an AST.

If you work at TypeSafe please try this.

Side note: This is probably how LLMs would perform with better encoders and next-latent prediction, so eventually those will beat this architecture out. Still amazing though.

An interesting example was the neomacs post on the frontpage yesterday.

https://neomacs.org

Some buttons on the front page don't even work.

And then you have Joachim Breitner using Claude Code to produce Soundness proofs in Lean4, mostly vibe coded.

It really is the QB in charge methinks

We really are entering the AI economy.

Now if only we knew if the stonks would go up or down (due to global turmoil) before I throw my savings at the SPY

Given how fast computers are these days there is probably not much overhead in having emacs run as the interpreter for an incrementally compiled language with HM types

I mean emacs has way more bugs than modern editors. Types would help you fix that and let you optimize code more since they help with "proving" that your code is correct.

I feel like emacs could be way better if e-lisp was typed but I write Rust for a living and Haskell/Lean for fun so maybe I'm biased

I guess I was thinking about the NFL but you're right about college.

Your point also extends the analogy, the AI might seem enough at lower levels due to that reason

I disagree that AI has broken the proxies.

Football analogy; AI is the wide receiver and the human is the quarterback. Even if you're the best WR in the game you're still not producing touchdowns unless you have a decent QB.

It's super easy to smell vibe coded projects.

All of these proofs and vibe code are impossible without human work. Call me when GPT whatever writes gcc from scratch

This cop is doing his part to get rid of flock

I actually couldn't have said it better. I'm brain rotted from LLM usage

You're assuming that the scripts are small.

e-lisp code is massive!

It's not nonsensical, but if you were writing emacs in rust, you hopefully would use a typed language instead of elisp

Watch demo button on the website doesn't work.

I mean gnu-emacs is already quite snappy in Wayland mode.

No point in reading anything AI related anymore. It's all slop.

We need to retvrn to rss feeds

They actually don't

The first time I have agreed with David Sacks

Yeah I use Astra Light at work. It's great but it probably cannot work on a project longer than a week without introducing architectural blunders