CLS

@ChengleiSi

Founding member @Recursive_SI & PhD @stanfordnlp | teaching language models to do research

Palo Alto, California

noviscl.github.io

Joined August 2018

4,107 Following

7,080 Followers

Replies Reposts Media

Pinned
CLS
@ChengleiSi
11h
I had tons of fun working on pretraining (& automating the pretraining) of Flint! It's a really good local model and a small preview of what we are cooking!
Richard Socher
@RichardSocher
12h
Introducing Flint, @Recursive_SI’s first model: a fast local coding model built with our automated research system, the Eureka Machine. • Best-in-class fast local coding • Up to 2x faster than comparable models: 125+ tokens/sec on a MacBook Pro, 150+ on RTX Spark • Built from
00:00
9
CLS
@ChengleiSi
11h
I’ve learned so much about pretraining from @yingbozhou_ai, a real legend
Yingbo
@yingbozhou_ai
11h
I've spent the past few months working on Flint. Excited to finally share our first model! Still a lot to improve, but that's also what makes this journey exciting.
Quote
Richard Socher
@RichardSocher
12h
Introducing Flint, @Recursive_SI’s first model: a fast local coding model built with our automated research system, the Eureka Machine. • Best-in-class fast local coding • Up to 2x faster than comparable models: 125+ tokens/sec on a MacBook Pro, 150+ on RTX Spark • Built from
1
CLS
@ChengleiSi
11h
everyone should follow @yihaocs, he's a legend 😉
Yihao Feng
@yihaocs
11h
Fun to build many things from scratch with the help of Eureka Machine :) More interesting things to come...
Quote
Richard Socher
@RichardSocher
12h
Introducing Flint, @Recursive_SI’s first model: a fast local coding model built with our automated research system, the Eureka Machine. • Best-in-class fast local coding • Up to 2x faster than comparable models: 125+ tokens/sec on a MacBook Pro, 150+ on RTX Spark • Built from
2
CLS
@ChengleiSi
11h
The vision here is cool, but I don't know how much I'm convinced by the bpb eval on internal repos. If "AI agents wrote over 99% of the code", then you are essentially comparing whose pretraining corpus contains more code written by the latest agents.
Behnam Neyshabur
@bneyshabur
12h
We formally announced Mirendil less than 4 months ago. Today, I'm excited to share an important milestone for: Pretraining at scale without a pretraining team. Mirendil's mission is to solve Recursive Self-Improvement (RSI) and use it to democratize frontier AI development,
4
CLS
@ChengleiSi
Oct 10
I had the fortune of listening to Jeremy Avigad discussing the future of mathematics in the age of AI, and I find his following slides deeply inspiring