@cong_ml

Automating Scientific Discovery @Recursive_SI. Prev: @GoogleDeepMind | @UBC_CS | @UniofOxford | @SakanaAILabs | @Waymo. The AI Scientist | Genie | SIMA | Gemini

 London, England
conglu.co.uk
Joined October 2019
  • Tweets832
  • Following1.8K
  • Followers4.9K
73Media
1/ From OpenAI's Navier–Stokes work to AlphaEvolve's faster algorithms, AI research is getting very real. Now, with Recursive Flint, we can experience that magic locally on our own laptops 💻 The whole research loop runs locally with a selection of tasks! Code + demos 🧵
17
36
11
311
40,509
1/ Can a model that runs on a laptop also drive GPU training research? Recursive Flint, our new 40B open model, improved NanoChat across 700+ experiments—finishing close to our historical Claude Haiku 4.5 results. We were pleasantly surprised. Here's what it tried 🧵
5
5
1
42
1,979
more replies
4/ Flint finished close to historical Haiku 4.5, with Sonnet 4.6 still ahead at the same experiment budget. Our 40B model uses just 2.5B parameters per token. We didn't expect it to get this far on a research task we'd associated with powerful proprietary models.
1
1
2
169
5/ Autoresearch helped us develop Flint. Seeing Flint contribute useful research itself is an encouraging early step. We're starting to explore how far that loop can go, and how to develop it safely. Meet Flint: xitter.cf/cong_ml/status/2112949… Try it locally: xitter.cf/cong_ml/status/2112966…
Cong Lu
@cong_mlOct 21
1/ From OpenAI's Navier–Stokes work to AlphaEvolve's faster algorithms, AI research is getting very real. Now, with Recursive Flint, we can experience that magic locally on our own laptops 💻 The whole research loop runs locally with a selection of tasks! Code + demos 🧵
1
3
170
Cool
Cong Lu
@cong_mlOct 21
1/ From OpenAI's Navier–Stokes work to AlphaEvolve's faster algorithms, AI research is getting very real. Now, with Recursive Flint, we can experience that magic locally on our own laptops 💻 The whole research loop runs locally with a selection of tasks! Code + demos 🧵
4
29
4,219
Nice mini repo in here for auto-research on your own machine. Auto-research can get spenny quickly, not anymore :)
Cong Lu
@cong_mlOct 21
1/ From OpenAI's Navier–Stokes work to AlphaEvolve's faster algorithms, AI research is getting very real. Now, with Recursive Flint, we can experience that magic locally on our own laptops 💻 The whole research loop runs locally with a selection of tasks! Code + demos 🧵
1
1
9
811
Cong Lu retweeted
Tim Shi
@timshi_aiOct 21
autoresearch on your local laptop! powered by Flint
Cong Lu
@cong_mlOct 21
1/ From OpenAI's Navier–Stokes work to AlphaEvolve's faster algorithms, AI research is getting very real. Now, with Recursive Flint, we can experience that magic locally on our own laptops 💻 The whole research loop runs locally with a selection of tasks! Code + demos 🧵
1
2
1
17
1,885
Cool application of @Recursive_SI's Flint for autoresearch—locally at home.
Cong Lu
@cong_mlOct 21
1/ From OpenAI's Navier–Stokes work to AlphaEvolve's faster algorithms, AI research is getting very real. Now, with Recursive Flint, we can experience that magic locally on our own laptops 💻 The whole research loop runs locally with a selection of tasks! Code + demos 🧵
1
31
2,773
Thanks to our Eureka Machine being put on mac inference + the model design, my MacBook barely makes a sound while churning flint :D
sacha🥝@alexUnder_skyOct 21
Really handy repository, thank you. Also, this repository shows how rich AI researchers are: a macbook might risk thermo throttling under heavy loads; and it costs like 3 grand. For such cool demonstration of unified memory, mac mini/studio seems like a better investement
1
1
16
732
Cong Lu retweeted
Yuandong Tian
@tydshOct 21
👉👉 AI Scientific Discovery Demos with our Recursive Flint-40B-A2B 🪨🪨!
Cong Lu
@cong_mlOct 21
1/ From OpenAI's Navier–Stokes work to AlphaEvolve's faster algorithms, AI research is getting very real. Now, with Recursive Flint, we can experience that magic locally on our own laptops 💻 The whole research loop runs locally with a selection of tasks! Code + demos 🧵
1
2
81
7,724
The automation of science will let us break all human priors. As we saw in math, the answer was often deceptively simple and under human noses all this time! 🫨
Thomas Wolf
@Thom_WolfOct 21
My bar for AGI is still an AI winning a Nobel. But an AI researcher looking at the lm_head (that *everyone* tells you to keep in high precision) deciding to quantize it to 4 bits AND being actually right (0.7pt from bf16) is definitely impressive. Even better when the result are open-weights pushed on the Hugging Gace hub. Congrats @Recursive_SI !
1
330
I was kind of shocked that autoresearch works as well as this with a tiny model like Flint
Cong Lu
@cong_mlOct 21
1/ From OpenAI's Navier–Stokes work to AlphaEvolve's faster algorithms, AI research is getting very real. Now, with Recursive Flint, we can experience that magic locally on our own laptops 💻 The whole research loop runs locally with a selection of tasks! Code + demos 🧵
2
7
127
11,765
5/ We used to associate useful autoresearch with the biggest closed, proprietary models. Flint can already propose ideas, write code, test it and improve results—all on a laptop. We were pleasantly surprised. You can inspect every experiment and choose what it works on next.
1
1
11
1,567
6/ Try it yourself :) We're sharing the code, evaluators, demos and all attempts, including failures. Give Flint your own task and a way to measure progress, and watch it run! Code + demos: github.com/conglu1997/local-… Meet Flint: xitter.cf/cong_ml/status/2112949…
Cong Lu
@cong_mlOct 21
1/ Introducing Flint, our first model at @Recursive_SI !! A blazing-fast 40B model on the local coding Pareto frontier, shaped by autoresearch 🔥 Built from scratch in under four months with our Eureka Machine and an incredible team. Here's how 🧵
1
1
28
1,772
Load more