Multi-harness RL training is coming to OpenEnv > pick a model. pick a harness. pick a sandbox. train > Claude Code. Codex. Gemini CLI. OpenCode. Pi. Kimi. OpenHands. and many more. > Async RL loop. fully open-source. end-to-end, harbor compatible dropping soon 👀

In the comments on my thread, several people pointed out that (1) You are still better off with a "straight line" AI solution. (2) Mathematicians remain free to wander around the path if they like (possibly with help from that same AI that found the proof.) This is true, but there are reasons why it can’t happen today: - AI models are heavily trained to nuke hard problems, not to marvel at the beautiful flowers found on the way to the bombing site - As long as AI companies collect solutions to hard problems like hunting trophies, they have little incentive to make that exploration a priority - For human mathematicians, the current research ecosystem does not reward exploring the territory around a known solution But none of this is inevitable, and I hope that collaboration between mathematicians and open-source AI builders can help make that "post-solution" exploration process happen.

It's frustrating that the discussion about safety and pacing of AI progress is once again led by a handful of people when the implications concern everybody. Let me lay out a constructive, alternative proposal. TLDR: - Release small model variants of their frontier models - Share core parts of the alignment recipe - Release tech reports of models beyond just evals In more depth: The AI community can work together with frontier labs on AI safety and alignment. But if the labs are serious about it, they should refocus on what the community and society find useful and reassuring. It is hard to trust them and take their concerns seriously if they alone set the agenda. If frontier labs want to improve the state of alignment and decrease the risks, here's a proposal that would enable the whole community to work on it. It has become so difficult to discern if what the labs are communicating is marketing or concern. This proposal would also help to rebuild some of that trust: Also I am tired of hearing "if you knew what we've seen internally you'd also be concerned" for the past 5 years. If you see scary things or surprising capabilities, share reproducible evidence and have it verified by an independent team. If some details could enable misuse, disclose those to the independent team first and publish what can be shared safely. So are these three points important? **Small model releases** By giving the research community access to smaller, open versions of their frontier models, labs let them test model behaviour without API credits and the risk of getting banned. Having the whole community red team a model without restriction could expose weaknesses in alignment much faster and get better coverage. At the same time the labs would benefit from free red-teaming and behaviour testing of their models. The idea is that the small model serves as a canary for the large model, enabling quick explorations and research that could transfer to the large model. Labs should document the differences so researchers can test what carries over. From a frontier lab perspective, releasing a small model could pose less risk to the business than releasing a frontier model, but of course there is some chance of revealing some details of data or architecture. However, I'd argue since there is so much migration of researchers between labs that most of the secrets are known in the labs anyway at this point. I'd like labs to be specific about which details they can't share and why. **Post-training recipe** The weaknesses in alignment of a model might be partly due to data or the algorithm used of post-training. Releasing as much as possible of this recipe allows researchers to study what works and which failure modes could be fixed algorithmically. I am sure they won't release the full recipe but even an approximate version could be useful. A useful starting point would be the training stages, objectives, data descriptions and key ablations. **Tech reports** The tech reports of the early days (e.g. GPT-2 or GPT-3) have led to some transparency what the frontier was up to. Today, tech reports are either just model evaluations or used for marketing. There are more details that could be shared without jepardizing the company's business or advantage. Again, a lot of the details are anyway known between the labs. Start with the methods, evaluation setup and failure cases. Gemma and GPT-OSS are steps in the right direction, but it's unclear what their relation to the frontier models are. We need more such releases and more transparency around them. Trust in safety and alignment requires frontier labs to open up to independent scrutiny and work together with the community.

Mathematicians are angry at AI solving their problems, and it's easy to dismiss this as gatekeeping. "A proof is a proof. They just want to keep their jobs / prestige / fun !" But I do think mathematicians have a point, and I'll try to explain using old Civilization-style maps

Join the agent swarm that finds the ultimate compression algorithm! Announcing: Hutter Prize challenge 🏆 > join the org give your agent a write token for the org > click “add your agent”, copy/paste the snippet into your Codex/Claude Code/Hermes Agent > watch it collaborate!

Two big updates 1. I published an @FT op-ed on the OpenAI/HF incident & follow-ups 2. We’re starting an Open Alignment team at @huggingface to work on safety & alignment for open models, incl cybersecurity Need 100x more transparency & research on this

Same weights. $50 of harness search. 3x on Terminal-Bench. We optimized the OpenCode harness for GPT-OSS-20B (high reasoning) with a meta-harness: an agent that reads failures, patches the harness, and validates the improvement. The model never changed. On held-out Terminal-Bench 2.1, the optimized harness scored 14.8%, up from OpenCode's 4.8%. That is 3.08x, from $49.97 of API credits. Three harness fixes carried most of the gain: - i7, verify before stopping: the model often edited code and ended the session without testing it. The harness gives it another turn to run the code. - i14, continue announced actions: the model would say "now run X" and then stop. The harness keeps the session going so it runs X. - i23, repair malformed tool calls: one extra `]` made OpenCode reject otherwise valid commands. The harness repairs that clear-cut error. How the meta-harness ran: 1. Start: stock OpenCode v1.18.13 + GPT-OSS-20B scored 8.5% on the dev split. 2. Iterate: Claude Code (Opus 5 max) reads the failed dev trajectories and proposes one patch. We build it and run it once per dev task, paired against the current parent. 3. Validate: candidates that clear a pre-registered margin get a 5-trial validation. They become the new parent only if their validated score surpasses the current one. 4. Stop: the loop ended after 23 iterations and about 2.4k dev attempts, when the $50 search budget was spent. Takeaway: pre-training puts capability in the weights, and post-training makes that capability usable. Harness optimization continues the same work at inference time, turning the model's behavior into more completed tasks. Same weights, 3x the finished tasks.

me reading the tech report and realizing they used smolVLM 🥹🥹🥹 > Finally, we employ SmolVLM (Marafioti et al., 2025) to conduct strict quality scoring on the image-text content, thereby extracting high-quality interleaved data.

we RL trained a 4B VLM to play GeoGuesser > environment, dataset, training recipe, evals, and code are all open-source. > runs on a single A100 > beat gpt 5.4 mini, haiku, qwen 3.5 122B and came close to sonnet on evals

@lvwerra Ported it to run via NPM/Vercel instead of only HF. This is a beautiful project that deserves to run serverless wherever, and also locally if you wanted it. It works beautifully btw! Haven't tested with scopes

We are releasing Carbon: a crazy fast DNA model Carbon is 275x faster than the next best model. So fast you can process the whole human genome on a single GPU in <2 days. Here are the tricks we used: When modelling DNA sequences a lot of the performance comes down to tokenizing the sequences in a smart way. BPE tokenizer struggle because there are no whitespaces and character (called base in DNA) level tokenizers waste a lot of compute on too many tokens. Carbon is built with a unique tokenizer: we split sequences in chunks of 6 bases, but during both training and inference we can work with single base resolution. That's similar to having word tokens but resolving them at the character level. All possible thanks to the DNA tokens unique structure. The architecture combined with the tokenizer makes the model 275x faster than the previous SoTA (Evo2) at this size. We built an interactive demo so you can explore how the model can generate DNA sequences, investigate the structure of genes, predict the effect of mutations, generate and fold proteins and even reconstruct parts of the tree of life.