


founder @commonwarexyz || previously @avalabs @coinbase @stanford || opinions are my own and will probably change


had a lot of fun formalizing Bajillion with @_patrickogrady
And on arXiv (thanks to @guruvamsip): arxiv.org/abs/2611.21883

And on arXiv (thanks to @guruvamsip): arxiv.org/abs/2611.21883

Bajillion is now in ALPHA (and under audit at @zksecurityXYZ): github.com/commonwarexyz/mon…

Introducing Claude Fable 5.2 and Claude Mythos 5.2, designed to tackle the most daunting technical challenges. They’re capable collaborators on ambitious projects in many fields, and show improvements in areas like coding, reasoning, and vision.

Nice DeepSWE score! Does it survive a lowercase filename? Introducing MaxxBench: 89 DeepSWE tasks with small changes to the instructions or starting code. Same engineering problem, slightly different requirement. Models clustered around 74% on DeepSWE end up 23 pp apart. Good at coding, or good at DeepSWE? 🧵

PSA on today's Reth / OP Reth releases A fix went out across the Reth ecosystem ensuring that all Reth-based nodes, DeFi bridges and state-proof backed systems are all safe. If you depend on Reth, please update to v.2.10.0, OP Reth v2.8.1, and Base version v1.5.3. We appreciate how quickly everyone moved to patch. This bug was identified by the Tempo team's fuzzing process, and Tempo is also patched. Reth’s legacy state root code had an edge case. For certain smart contract storage layouts, a carefully crafted transaction could make it produce the wrong state root. Reth’s block builder uses this legacy trie calculator on lower-end hardware and on OP Stack chains, so those setups were exposed. L2s like OP Stack chains or Base are unaffected because this could not impact their withdrawals system. Ecosystem bridges do not use state roots, they use receipt roots or multisignatures, so they are unaffected by this. Reth is critical infrastructure for the entire EVM ecosystem. We take it very seriously. And we are sorry for the inconvenience caused.

The only moat left for software will soon be "economies of scale" (how cheap can you acquire the infra to run it). ~network effects/licenses will fall to 0 with agents strip mining the internet for the best value (and building whatever they need along the way to do so).


Some news: I’m joining OpenAI to lead product management for Core Products! I have a lot of respect for OpenAI for showing the world that scaling works, for bringing AI to over a billion people, and for believing great products are core to the mission. I’m excited to work with @thsottiaux, an incredible PM team, and everyone across OpenAI. Together, we’ll build agents that have your back and your team’s, and keep you in control, so you get more time for the people and work you love. Lots to learn, and lots to build. Please keep your feedback coming!

We asked GPT-6 Astra to edit Qwen3.5-0.8B weights to improve math. It directly changed only 0.016% of weights, fixing 0.8B's reasoning loops & improving math up to 6x and even instruction following 3x, without gradients or training. We call this Weight-Space Programming: frontier LLMs can now natively read/write neural network weights! 🧵 Blog: rikiyat.github.io/blog/weigh…

I just built a *functioning* React app by describing what I wanted to GPT-3. I'm still in awe.


We are of a potential security incident related to Binance. As a reminder all collateral with exposure to exchanges resides in off- exchange custodians, not directly with the exchange themselves. As we learn more details on the incident we will provide updates here.

The blockchain products launching over the next few months will blow you (and your agent's) mind. AFAICT web3 tx cost/latency (w/ verifiable finality) will be less than or equal to web2 api cost/latency by the end of 2026.
