rahen

[rahen @rahen_ml]

@mSykeCodes Can't be worse than this one...

[rahen @rahen_ml]

@amishjt @fchollet LeCun's Lush, maybe? Lisp is very effective for neural networks.

[rahen @rahen_ml]

@schteppe I doubt anything will ever beat C when it comes to performance. What you want is C written by an LLM with strong safety guidelines.

[rahen @rahen_ml]

@bratton Now tell them 'ed' was the standard editor. On teletypes, not terminals.

[LaurieWired @lauriewired]

There’s a famous Usenet story about a programmer (Mel) who refused higher level abstractions.

It was the late 1950s, and even in that era, Mel was…well today we’d call him a boomer.

Mel only wrote in raw hexadecimal. He didn’t approve of compilers, and refused to use optimizing assemblers.

"You never know where it's going to put things”, he said.

Everyone else in the company was moving on to FORTRAN, and they didn’t understand why Mel was so stubborn about using new tools. He *loved* self-modifying code.

“If a program can’t rewrite its own code”, he asked, “what good is it?”

Mel eventually left the company, and other engineers were tasked with understanding what was left.

Mel’s hand-optimized routines always beat the assemblers; but some of it looked absolutely bizarre.

One engineer took ~2 weeks to understand why there were loops with no exit condition…yet the program worked fine.

I won’t spoil all the details, you should really read it, it’s short. But it’s a fantastic piece on “what defines a real programmer?”…which is becoming increasingly relevant in this vibe-coded era.

I strive to understand computers as deeply as Mel! If we aren’t careful, we’re going to lose the “Mels” of this world to time.

That’s part of why I go so deep in my youtube videos. I hope that younger viewers are genuinely fascinated by the inner workings of our machines, instead of handing everything off to higher abstractions.

[rahen @rahen_ml]

@JonathanRoss321 In an age of agentic development pouring oceans of unreviewed code every day, the promise of saying no has already become the value proposal. I want bloat removers, not slop zombies.

[rahen @rahen_ml]

@MIT_CSAIL @a1zhang Love the VT-100! Is it there for a specific reason?

[rahen @rahen_ml]

@AlexRozinov @ylecun Ditto. A modern Transformer in Lisp clearly has a mathematical beauty lacking from the noisier, imperative Python-based code.
https://t.co/mtYETallHG

Unfortunately Lisp remains a niche in ML nowadays.

[rahen @rahen_ml]

@fleetwood___ Don't do sequential training. The weights should be trained on all tasks simultaneously.

[rahen @rahen_ml]

@Mayhem4Markets LLMs are hitting their limits. MOE, KV caches, quantization and compactions can only go so far, and parameters don't scale well.

What Anthropic needs is an architectural breakthrough, not more parameters.

[rahen @rahen_ml]

@bugakovas @davepl1968 Yes, the source code is already available :-)
https://t.co/LivBskukNV

[rahen @rahen_ml]

@davepl1968 Very nice! Maybe should you also mention the original author, Damien Boureille:
https://t.co/770QxLSzUL

[rahen @rahen_ml]

@draganrocks Why something like this gets 7 "likes" and not 50K is beyond me...

Beyond GPUs, Lisp and machine learning are a match made in heaven.

[rahen @rahen_ml]

@DmytroKrasun I'm working on a functional ML framework and have to ship it with a context file, because LLMs constantly try to write models "the PyTorch way".

[rahen @rahen_ml]

@jsuarez Time to go back to Lisp!
https://t.co/nx6i8Io0rI

NanoGPT in Sheaf is just 120 lines long:
https://t.co/mtYETallHG

[rahen @rahen_ml]

@yassineyousfi_ Or ditch Python altogether...
https://t.co/MLzZ4jX08X