#compchem Good read: Physics-based molecular fingerprints from spectral graph theory provide efficient geometry-aware measures of chemical similarity.
https://t.co/azO2mDVPNa
Thank you to Grant Sanderson @3blue1brown for having me on his podcast – and I hope you enjoy our open-hearted, tentative, exploratory conversation about math in the age of AI. We tried out a lot of ideas on each other. https://t.co/kDTQWcvp6l
I did say we'd be releasing our models periodically...introducing Gemini 4 Carbon, our next frontier model.
Following Argon's release just two weeks ago, Carbon shows the pace at which we’re pushing the frontier, it ranks #1 on the recently updated v5.0 of @ArtificialAnlys Intelligence Index at 59. It also leads across key cybersecurity evaluations scoring 88.2% on SEC-bench Pro for finding vulnerabilities, 54.8% on ExploitGym for exploitation, and 71.0% on CWE-bench for patching.
We’re seeing significant improvements in long-horizon tasks even compared to Argon, and we’ve strengthened frontier safeguards.
As we did with Argon, we’re going to roll it out gradually + responsibly. Some trusted cyber defenders will get access starting this week. We'll be expanding it to more partners through our Fairwind Program, ahead of wider availability soon, so hang tight!
Starlink is licensed in over 165 countries and has spent five years complying with every single law and requirement of the government of India, so why still no license?
Is Ambani the real boss of India?
Unfortunately, we are being blocked by certain oligarchs in order to maintain their monopolistic chokehold on the Indian people. You can guess who they are …
This is a crime against the people of India!
If you’re a researcher who uses Claude Code or Codex for your daily work, consider using open models instead.
Recent events have shown why owning the entire stack is so important. While OpenAI and Anthropic currently offer the strongest models, using them means working on their terms. OpenAI and Anthropic have shown they are not afraid to alter model output, service, and access if user interests conflict with their business interests.
OpenAI:
- Sep 2026: accused of using Codex data from Buckmaster and Alpöge to race to a solution to Navier-Stokes using their massive compute advantage. OpenAI later admitted that they “cannot rule out that de-identified data derived from their usage of our products helped improve our models”.
- Aug 2026: announced removal of all OpenAI models on Cursor after their SpaceX acquisition
- Dec 2025: injected ads into ChatGPT conversations, even for users paying $200/mo subscriptions
Anthropic:
- Jun 2026: launched Fable 5 with safeguards that limit Claude’s effectiveness at ML research tasks through interventions that are not visible to the user
- Apr 2026: removed subscription coverage for third-party tooling such as OpenClaw and Pi
- Jan 2026: cut off xAI engineers’ Claude access in Cursor
The only way to protect yourself from these kinds of interventions by the labs is to own the model, the tooling, and the data. This is particularly important for researchers, who often work on confidential projects and with sensitive data.
You should do research on your terms, with tools you control and work that remains yours.