The best ChatGPT that $100 can buy.

Beads - A memory upgrade for your coding agent

DDN: A novel generative model with simple principles and unique properties. (ICLR 2025)

Code for Words That Make Language Models Perceive

An alignment auditing agent capable of quickly exploring alignment hypothesis

🌾 OAT: A research-friendly framework for LLM online alignment, including reinforcement learning, preference learning, etc.

A paint and ASCII art application for the terminal.

What does gpt-oss tell us about OpenAI's training data?

personal site

AT application for ESP32/ESP32-C2/ESP32-C3/ESP32-C6/ESP8266

The Personality Illusion: Revealing Dissociation Between Self-Reports & Behavior in LLMs.

Robin: A multi-agent system for automating scientific discovery

Stochastic Parameter Decomposition

GLM-4.5: Agentic, Reasoning, and Coding (ARC) Foundation Models

🧠 Cognitive load is what matters

Inspect: A framework for large language model evaluations

A minimal SAE implementation ready for hacking

Code for reproducing our paper "Not All Language Model Features Are Linear"

Paper2Agent is a multi-agent AI system that automatically transforms research papers into interactive AI agents with minimal human input.