4 AI Numbers that Surprised Me This Week

A new world for RL, benchmarks underestimating AI ceiling, an interesting but obscure puzzle, and a new world of free (?) intelligence.

OpenAI Code Red + Claude 'Soul' Document

What if artificial intelligence really does become a bit too cheap to meter?

Kimi K2-Thinking Simple Bench + AI Explained app

An early insight into this fascinating new model, and a new curated benchmarking hub/app.

Gemini Deep Think Revelations + Ilya Sutskever's 'AGI Bunker'

o3 in 2 weeks and Former OpenAI Researcher Predicts 2027 ASI

Anthropic CEO: "Beat China with Self-Improving AI" - no red flags there - plus Google's Titans

Is it me or is the world getting a little more cuckoo every few months?

o3 steals the spotlight, but Google shipped some serious goods

OpenAI Fights ‘Simple’ with ‘Simple’ + Contradictory Leaks + Dev Day

The 3rd Era of AI Language Models

OpenAI’s Orion ‘Flagship LLM’

OpenAI ‘Strawberry’ and What-Bench?

NSA Joins the Conversation and New $1m AGI Prize to Test the I in AI