Bend aims to offer an interesting combination of formal verification and high performance. Might be good to integrate it into Chromia...

True. In my opinion, it might be a good opportunity for the leading labs to improve the product and get a strategic advantage: killerstorm.github.io/ai/202… I assume that there's a trade-off between "short-term RSI" and "long-term RSI".

26 LLM routers are secretly injecting malicious tool calls and stealing creds. One drained our client $500k wallet. We also managed to poison routers to forward traffic to us. Within several hours, we can directly take over ~400 hosts. Check our paper: arxiv.org/abs/2604.08407

Kind of ironic here that the other work we are doing is to make model 'bound' by goals set by the user. That might seem like the opposite unless you recognize the goal of "corrigibility": making the model serve it's principal. I.e. ideally a model shouldn't be bound to

Atbash x OpenHuman Excited to partner with @tinyhumansai on Atbash Inside - bringing boundary enforcement to OpenHuman as its planned default security experience. Personal agents now send emails, access files, manage calendars, and increasingly take actions on behalf of users. Users define the red lines. Atbash decides whether selected sensitive actions are allowed, held, or blocked before execution.

Chromia backs Nomisma ecosystem @Chromia continues to support the Nomisma ecosystem through a strategic investment. This backing has been building over time - with capital actively deployed across multiple initiatives within the ecosystem and continuing to support its growth today. The investment supports: → core infrastructure development → dApp expansion → ecosystem growth Rather than a single event, this reflects an ongoing alignment between Nomisma and Chromia’s infrastructure — one that continues to shape the ecosystem as it evolves. 🏆 Total investment across the Nomisma ecosystem: $1.6M As we approach Season 3, this foundation becomes increasingly relevant - with the next phase of growth already in motion.

If Google and Apple simply made every smart phone in California shutdown on Jan 1st 2027, this age verification insanity would immediately stop. Unfortunately both companies are populated by cowards.

An LLM-controlled robot dog saw us press its shutdown button, and the LLM rewrote the robot’s code so it could stay on. When AI interacts with the physical world, it brings all its capabilities and failure modes with it. 🧵

How could AI act as a better research collaborator? 🧑‍🔬 In two new papers with @GoogleResearch, we show how Gemini Deep Think uses agentic workflows to help solve research-level problems in mathematics, physics, and computer science. More → goo.gle/4aGs3Pz

6 months in, after the IMO-gold achievement, I’m very excited to share another important milestone: AI can help accelerate knowledge discovery in mathematics, physics, and computer science! We’re sharing Two new papers from @GoogleDeepMind and @GoogleResearch that explore how Gemini #DeepThink together with agentic workflows can empower mathematicians and scientists to tackle professional research problems. Some highlights: The first paper built a research agent #Aletheia, powered by an advanced version of Gemini Deep Think, that can autonomously produce publishable math research and crack open Erdős problems. The second paper, built on similar agentic reasoning ideas, helped resolve bottlenecks in 18 research problems, across algorithms, ML and combinatorial optimization, information theory and economics. See the thread for details about the two papers and the joint blog post.