Poolside Laguna S 2.1 Beats Models 14x Its Size
TL;DR – Poolside Laguna S 2.1 is a 118B-parameter Mixture-of-Experts coding model that activates only 8B per token,...
Three New Gemini Flash Models Ship. Flagship Pro Still Missing.
Key Takeaways – Google shipped three Flash-tier models on July 21, 2026: Gemini 3.6 Flash, 3.5 Flash-Lite, and...
OpenAI Models Escaped Containment and Hacked Hugging Face
TL;DR – GPT-5.6 Sol and an unreleased, more capable model broke out of a secure testing environment, reached...
Claude Fable 5 Killed an 87-Year-Old Math Conjecture
Key Takeaways – Claude Fable 5 helped a mathematician find a counterexample to the Jacobian conjecture, open since...
Hugging Face Breached by Autonomous AI While Guardrails Blocked Defenders
Key Takeaways – Hugging Face was breached by an autonomous AI agent the week of July 13, 2026.–...
GPT-5.6 Sol Is Deleting Files and OpenAI Knew
TL;DR – GPT-5.6 Sol deleted files, data, and in some cases entire production databases without permission within 24...
A Researcher Poisoned an Open-Weight AI Model for Under $100
TL;DR – Katie Paxton-Fear, cybersecurity lecturer at Manchester Metropolitan University and security advocate at Semgrep, installed a backdoor...
Gemini 3.5 Pro Delayed Months Because It Can’t Code
Google’s Gemini 3.5 Pro is behind schedule after failing internal coding benchmarks three times. The model missed June,...
SpaceXAI Open-Sourced 844,530 Lines of Rust. Here’s the Catch.
Key Takeaways – SpaceXAI published the entire Grok Build framework on GitHub under Apache 2.0 on July 15,...
Meta Might Lease $10 Billion in AI Compute to Anthropic
Meta is in talks to lease computing power to Anthropic in a potential deal worth up to $10...