Semantic ifs from open models, on a 3090 at home. Independent; not affiliated with Jev or TypeSafe.

Message Passing Neural Networks for Molecule Property Prediction

Exo is an agent + harness architecture that is fully recursive, able to safely edit all aspects of itself at runtime to get better at your tasks.

100 days of LLM inference engineering — daily posts, experiments, and visualizations

Materialize is a program for converting images to materials for use in video games and whatnot

Which host should serve your open model? Run your workload across pinned, verified providers and get one table: success, cost, latency, cache hits.

My reasearch of losslessly compressing LLM weights.

Reliable, minimal and scalable library for pretraining foundation and world models

Open-source, developer-first PaaS: git push → live HTTPS URL on hardware you own, including a Raspberry Pi behind CGNAT, via a zero-trust cloud relay.

Un-0: an image generator powered by a simulated system of coupled oscillators, an example of an emerging physical computing substrate.

A reimplementation of Walter Fontana's Algorithmic Chemistry (AlChemy)

Tiny Model, Big Logic: Diversity-Driven Optimization Elicits Large-Model Reasoning Ability in VibeThinker-1.5B

Monorepo & docs

An interface library for RL post training with environments.

Open-source framework for the research and development of foundation models.

Open-source evolutionary coding agent: LLMs discover and optimize algorithms through iterative evolution. Compounding memory, adaptive model selection, multi-language.

Agentic RL on Any Harness at Scale

We propose Bidirectional Evolutionary Search (BES), a search framework that couples forward candidate evolution with backward goal decomposition.

Unquestionable PDDL 3.1 parser

Claude Code skill that turns 'make me a chart' into a Tufte-compliant chart. Distilled from Edward Tufte's three foundational books.

SR²AM: Efficient Agentic Reasoning Through Self-Regulated Simulative Planning

Method for Long Context RLMs using verifiable Lambda Calculus

Google's open agentic orchestration runtime

HRM-Text is a 1B text generation model based on the HRM architecture, strengthened by task completion and latent space reasoning.

🪨 why use many token when few token do trick. Viral skill + proxy for coding agents that cuts 65% of tokens by talking like a caveman.