Loading...

Tag trends are in beta. Feedback? Thoughts? Email me at [email protected]

Pacing the Frontier is not the actual goal for AI labs

Watch AI materials-science and bioscience abilities closely

How My Students Think About AI

Astra and Fable still hack on simple variants of alignment evals from 2025

What is nueralese and why is it bad

How concerned should we be about Astra's recurrent architecture?

Misaligned AIs could use killer robots to take over

What I did in the hedonium shockwave, by Emma, age six and a half

Don't Build Mindreading

Why I Left Google DeepMind

Unless Its Governance Changes, Anthropic Is Untrustworthy

An OpenAI model left notes about how to evade containment; we need more details

GPT-5 writing a Singularity scenario (2025)

Google Can't Math Parsecs

Intuitive Self-Models (2024)

Even "illegible" Mythos reasoning traces seem pretty legible

Probabilistic, Reformative Justice

Go Players Disempower Themselves to AI

Opus 4.7 Part 3: Model Welfare

Working hurts less than procrastinating, we fear the twinge of starting (2011)

Most people can't juggle one ball

The effects of caffeine consumption do not decay with a ~5 hour half-life

We're running out of benchmarks to upper bound AI capabilities

The AlphaFold moment for materials is not any time soon

You can't imitation-learn how to continual-learn

I am definitely missing the pre-AI writing era

The optimal age to freeze eggs is 19

LessWrong Policy on LLM Use

To the Polypropylene Makers

Gemma Needs Help

More →