Henry Lu

@HenryL_AI

San Francisco, CA

Joined June 2016

175 Following

792 Followers

Replies Reposts Media Articles

Pinned
Henry Lu
@HenryL_AI
Oct 20
Can we train language models to propose better ideas to improve LM—not just run them? Introducing 𝐋𝐌𝟐𝐋𝐌: Language Models Learn to Improve Language Models. We train a 27B model with RL to propose improvements to language models. On held-out autoregressive and diffusion LM
3
Henry Lu
@HenryL_AI
Nov 2
Exciting, shall we scale up LM2LM to more model training tasks, and build the best open-weight autoresearch models for model training? arxiv.org/html/2610.2334…
Chenru Duan
@chenru_duan
Nov 2
We’re introducing Hydrogen-1.0, our first open-weight model for scientific research. We post-trained GLM-5.3 (@Zai_org) on our internal scientific data, lab records, and discovery trajectories. It scores 38.3% on FrontierScience-Research vs GPT-6-Astra’s 33.3% and scores 93.0%
Henry Lu
@HenryL_AI
Oct 23
The iconic RSI model launch: One researcher + agents pretrained a 100B–500B model in 3.5 months. Agents wrote over 99% of the code. But the code percentage doesn’t measure the remaining bottleneck: deciding which experiments are worth running. I broke down what the agents
Henry Lu
@HenryL_AI
Oct 22
Article
Recursive's Flint and Mirendil-Base: What Did the AI Actually Do in Building Frontier Models?
Breaking down Flint and Mirendil-Base: what the agents actually did, how the research worked, and what comes next. What “AI-built” actually means in these two releases This week, @Recursive_SI and
Henry Lu
@HenryL_AI
Oct 22
What did the agents actually do in Flint and Mirendil-Base? @Recursive_SI’s Flint release describes specific architecture, quantization, and kernel changes found by Eureka. @mirendil reports that one researcher + agents pretrained a 100B–500B model in 3.5 months, with agents
Henry Lu
@HenryL_AI
Oct 22
Article
Recursive's Flint and Mirendil-Base: What Did the AI Actually Do in Building Frontier Models?
Breaking down Flint and Mirendil-Base: what the agents actually did, how the research worked, and what comes next. What “AI-built” actually means in these two releases This week, @Recursive_SI and
Henry Lu
@HenryL_AI
Oct 22
Article
Recursive's Flint and Mirendil-Base: What Did the AI Actually Do in Building Frontier Models?
Breaking down Flint and Mirendil-Base: what the agents actually did, how the research worked, and what comes next. What “AI-built” actually means in these two releases This week, @Recursive_SI and

Henry Lu

Henry Lu’s posts