Stories by frozenseven

Dreamer 4 - Training Agents Inside of Scalable World Models

AI Companies Data Hub

The Bitter Lesson's Bitter Lesson

The QMA Singularity

Beyond Power Laws: Scaling Laws for Next-Token Prediction

Why GPT-5 used less training compute than GPT-4.5 (but GPT-6 probably won't)

Anything Max – an autonomous software engineer

Euclyd – Startup to Take on AI Inference with Sip, Custom Memory

Release: Power Retention

Using GPT-5 to prove new theorems on matrix multiplication

Moe at Scale: Making Sparse Models Fast on Real Hardware

[ARC-AGI-2 SoTA] Efficient Evolutionary Program Synthesis

Information Flows Through Transformers

Gauss, an Agent for Autoformalization

The Training Imperative

AI Quests: Bringing AI literacy to the classroom

FinePDFs: Liberating 3T of the finest tokens from PDFs

If We Build It, Everyone Lives

Deathwatch – Archive Team

Leonardo Torres Quevedo

1
2
3