LLMs

Saturated Benchmarks Are Not Solved Benchmarks

I Sense a Contradiction

Autoregressive Errors are Independent?

LLM Architecture Scales

Stacking at Length

Grammatically Correct

o3 is Coming

LLMs as a Repository of Vector Programs

AGI

Matching Robots

Why Autonomous Software?

Maintaining Old Work

AGI as the Best Gamer

How to Evaluate Reasoning Capabilities

What is Reasoning?

Learning Efficiently

A Basic Framework for General Intelligence

GPUs are Awesome

Another Great Year for AI, 2024 -> 2025

LLM

AGI as the Best Gamer

RL

Matching Robots