Moisés Arizpe

[Moisés Arizpe @_numerico]

Worked on a quick implementation for Apple Silicon here https://t.co/SK5C4Q3AnQ

[Nick Levine @status_effects]

New work with @AlecRad and @DavidDuvenaud:

Have you ever dreamed of talking to someone from the past? Introducing talkie, a 13B model trained only on pre-1931 text.

Vintage models should help us to understand how LMs generalize (e.g., can we teach talkie to code?). Thread:

[Moisés Arizpe @_numerico]

@Heliouz__ Buenas gráficas! Interesantemente se observan ahora dos modas.

[Moisés Arizpe @_numerico]

@SemiAnalysis_ You are assuming their LLM tokens are words. They might be transactions.

[Moisés Arizpe @_numerico]

@pablo_majluf IQ es irrelevante si alguien es ideológico.

[Moisés Arizpe @_numerico]

@karpathy @LinghuaJ Map Reduce on docs sounds like this https://t.co/pffyvQ7zBb

[Moisés Arizpe @_numerico]

@errantestrella @The_xx Hola! Ya no los cambio por los del domingo, pero te puedo vender uno del viernes. Te interesa?

[Moisés Arizpe @_numerico]

@karpathy Have you tried ChatGPT Pro as chief scientist?

[Moisés Arizpe @_numerico]

@VictorTaelin I think we just have to reverse engineer GPT Pro

[Moisés Arizpe @_numerico]

@SIGKITTEN =0====00

[Moisés Arizpe @_numerico]

@fchollet I think it’s more like going from a manufacturing factory to a chemical processing plant.

[Matthew Honnibal @honnibal]

I still see a lot of people discussing LLMs as next-token predictors, which is by now quite a misunderstanding. A related opinion is that LLM progress will probably plateau. This post explains why I don't think the "plateau" argument holds up. https://t.co/fJPBoWs2aX

[Moisés Arizpe @_numerico]

@fchollet Excalidraw mcp

[Moisés Arizpe @_numerico]

@moaimx CSV?

[Moisés Arizpe @_numerico]

@chrisalbon @infinitehumanai Tell it to create reports with charts that let you explore inputs and outputs. Also to include code snippets etc. just as you would do with a direct report. The inspection layer does not need to be code.

[Moisés Arizpe @_numerico]

@8teAPi Full i/o to the digital world is now available.

[Moisés Arizpe @_numerico]

@schickling You need to see only the state variables of code.

[Moisés Arizpe @_numerico]

@sama

[Moisés Arizpe @_numerico]

@simonw @geoffreylitt “The team did it”

[Lee Robinson @leerob]

I migrated cursor​.com from a CMS to raw code and Markdown.

I had estimated it would take a few weeks, but was able to finish the migration in three days with $260 in tokens and hundreds of agents.

Here's how I did it + all my my usage stats.

https://t.co/QIAOmLsffx

[Noam Brown @polynoamial]

Today, we at @OpenAI achieved a milestone that many considered years away: gold medal-level performance on the 2025 IMO with a general reasoning LLM—under the same time limits as humans, without tools. As remarkable as that sounds, it’s even more significant than the headline 🧵