We built @nvidia OpenShell out in the open with a commitment to keeping it open source. Our research is open, too.
Learn more about OpenShell Research in my recent Dev Note: https://t.co/OTBKsJ3Nje
There is a made-up letter circulating online asking NVIDIA’s board to reinstate Jensen, claiming my involvement. This letter is obviously fake, and even if it were real I would not sign it.
I am grief-stricken that Jensen is no longer with NVIDIA. At the same time, I fully support NVIDIA’s board and have confidence in their decisions.
I’ve worked with Colette for years and know her as an extraordinary leader. She has my full support.
This whole thing sucks, and I wish none of it had happened. It’s a really sad day!
But I want to make one thing clear: we’ll keep fighting for open-source AI, harder than ever. This is too important to slow down now. Let’s go!
OpenShell is now part of the CNCF Sandbox. Originally developed and contributed by @NVIDIA, the project brings community-led governance and infrastructure-level boundaries to autonomous AI agents.
Read the perspective from Jonathan Bryce (@jbryce) and Ali Golshan (@a_golshan) on securing agent runtimes: https://t.co/N09d9Z2Yr6
Explore the project: https://t.co/7zM18PpwhT
We built @nvidia OpenShell out in the open with a commitment to keeping it open source. Our research is open, too.
Learn more about OpenShell Research in my recent Dev Note: https://t.co/OTBKsJ3Nje
This is a big moment for NVIDIA and the OpenShell team. I’m super proud of what we have built and look forward to continue building with the CNCF community 💜
Securing AI agents starts at the runtime layer, not just in prompts.
OpenShell is now in the CNCF Sandbox, bringing policy isolation, network controls, and credential management to autonomous agents.
Read more: https://t.co/QaFxfAGfYf
#OpenSource#AI
This is a great thread. We’re thinking a lot about formal verification on the OpenShell Research team. Defining formal semantics for actions agents can take over time is an important and challenging problem indeed.
Deciding how to help non-expert humans engage with formal verification specifications seems like one of the more important non-verifiable UX problems facing the rise of formal verification. 🧵
I have recently had more time to build and have been merging ~10s of PRs per day. It’s amazing how much agentic-development wisdom Lauren packed into this post on pstack.
@jerryjliu0 Highly recommend giving @t3dotcodes a try. Best remote experience, and you can seamlessly switch between harnesses in the same awesome UI. Not to mention cross-harness subagents. Seriously it rocks!
From Opus just now: "Fair challenge. I described those workarounds from the comment in upstream.toml without checking them myself."
You see this is why @poteto banned code comments.
@Vtrivedy10@t3dotcodes agreed! optimizing for the task when you can is for sure the way to go.
honestly all the main players have become so good that I often simply reach for the best, most flexible UX to get the job done.
Introducing Samaya Sterling, our new model post-trained for challenging financial research that outperforms the frontier! Sterling started off as a two person research effort to see if we could increase efficiency in agentic orchestration. To our surprise we were able to post-train a much smaller model (Nemotron 3.5 Lightning, a 30B mixture-of-experts model with 3B active parameters) to greater accuracy (58.9% vs 55.9% on FrontierFinance) and 11x lower cost ($0.61 vs $6.89) than Claude Fable 5.1, the best performing frontier model.
We did this via a four-stage training process, two stages of SFT where the model first learns to research with our financial tools and how to construct a good agentic research strategy. The second two stages are RL, with an offline GRPO stage where the model further hones its research strategy from eight teachers that provide diverse approaches, and an online GRPO stage where the model can make end to end attempts and learn from its mistakes.
We see substantial gains from the co-adaptation of the model and harness through post-training, both through the model learning to effectively use the high-quality financial tools, and better compaction approaches that let even a small model execute on longer and deeper research.
We are exploring multiple follow-up directions, from multi-agent research swarms using this small model, to training firm-specific models to encode unique judgement, to training a larger model, where early results show even greater gains from the training process.
The full research is linked below!
@MatthewBerman 💯!! T3 has been a VS Code moment for me. Been using it for a couple weeks and it’s been a huge unlock. Best coding agent app experience I’ve had.