Yahya Jabary

dreamerv3. Mastering Diverse Domains through World Models

tdmpc2. Code for "TD-MPC2: Scalable, Robust World Models for Continuous Control"

mpmath. Python library for arbitrary-precision floating-point arithmetic

cax. Cellular Automata Accelerated in JAX (Oral at ICLR 2025)

omni-epic. OMNI-EPIC: Open-endedness via Models of human Notions of Interestingness with Environments Programmed in Code (ICLR 2025).

pulp-runtime. Simple runtime for Pulp platforms

colibri. Run GLM-5.2 (744B MoE) on a 25GB-RAM consumer machine — pure C, zero deps, experts streamed from disk. Tiny engine, immense model. 🐦

pxpipe. cut Fable 5 token usage by rendering text context as images

mjviser. Web-based MuJoCo viewer powered by Viser

mjlab. Isaac Lab API, powered by MuJoCo-Warp, for RL and robotics research

mink. Python inverse kinematics based on MuJoCo.

mujoco_warp. GPU-optimized version of the MuJoCo physics simulator, designed for NVIDIA hardware.

warp. A Python framework for GPU-accelerated simulation, robotics, and machine learning.

lean-eureka. EURISKO-inspired reflection, LCF-style trust — a fixed-gate verified discovery system in Lean 4

jax_privacy. Algorithms for Privacy-Preserving Machine Learning in JAX

reaf. Robotics Environment Authoring Framework

deepmind-research. This repository contains implementations and illustrative code to accompany DeepMind publications

dm-haiku. JAX-based neural network library

acme. A library of reinforcement learning components and agents

mujoco_mpc. Real-time behaviour synthesis with MuJoCo, using Predictive Control

thinksound.cpp. Standalone C++/GGML runtime for ThinkSound text->sound-effect generation

awesome-secure-program-synthesis. vibecoding, but correct and safe and secure and

axiomander. Code by Contract Dev Environment

DeepSeek-OCR. Contexts Optical Compression

formal-disco. A framework for scalable, distributed discovery systems.

peano. An environment for learning formal mathematical reasoning from scratch

dynamatic. DHLS (Dynamic High-Level Synthesis) compiler based on MLIR

basedpyright. pyright fork with various type checking improvements, improved vscode support and pylance features built into the language server

lemmascript-seal. The prover proves the present; the seal protects the past. An append-only freeze for verified contracts.

breaktimer-app. Manage periodic breaks. Avoid eye-strain and RSI.

lerobot. 🤗 LeRobot: Making AI for Robotics more accessible with end-to-end learning

DynamixelSDK. ROBOTIS Dynamixel SDK (Protocol1.0/2.0)

mujoco. Multi-Joint dynamics with Contact. A general purpose physics simulator.

mujoco_menagerie. A collection of high-quality models for the MuJoCo physics engine, curated by Google DeepMind.

tenferro-rs. A Rust-native tensor & autodiff stack for scientific computing

vellvm. The Vellvm (Verified LLVM) coq development.

llm-verifier-interface. Based on lectures on the LLM-Verifier Interface at the Summer School on Foundations of Programming and Software Systems (FoPSS 2026)

Formalization-High-Girth-High-Chromatic-Number. Lean formalization of the mathematical proof that there exist graphs with arbitrarily high girth and chromatic number.

z3. The Z3 Theorem Prover

ponytail. Makes your AI agent think like the laziest senior dev in the room. The best code is the code you never wrote.

lean4-mlir. Lean specification of neural architectures with verified IREE codegen.

HRM. Exploration into the proposed architecture from Sapient Intelligence of Singapore 🇸🇬

tiny-recursive-model. Unofficial implementation of Tiny Recursive Model (TRM), improvement to HRM from Sapient AI, by Alexia Jolicoeur-Martineau

AccelEval. Benchmarking and Profiling LLM for accelerating CPU code

coconut. Training Large Language Model to Reason in a Continuous Latent Space

LeanProject. A template for blueprint-driven formalization projects in Lean.

mathlib4. The math library of Lean 4

format_lean. A Lean file formatter

timesfm. TimesFM (Time Series Foundation Model) is a pretrained time-series foundation model developed by Google Research for time-series forecasting.

CADAM. CADAM is the open source text-to-CAD web application

cutile-rs. cuTile Rust provides a safe, tile-based kernel programming DSL for the Rust programming language. It features a safe host-side API for passing tensors to asynchronously executed kernel functions.

qlib. Qlib is an AI-oriented Quant investment platform that aims to use AI tech to empower Quant Research, from exploring ideas to implementing productions. Qlib supports diverse ML modeling paradigms, including supervised learning, market dynamics modeling, and RL, and is now equipped with https://github.com/microsoft/RD-Agent to automate R&D process.

ExoBLAS. BLAS Library in Exo

ASCILINE. A high-performance, real-time ASCII video rendering engine. Streams binary-encoded frames via WebSockets for ultra-low latency, 30 FPS playback using HTML5 Canvas and requestAnimationFrame.

IsaacLab. Unified framework for robot learning built on NVIDIA Isaac Sim

quadrants. High-performance multi-platform compiler for physics simulation

OpenBLAS. OpenBLAS is an optimized BLAS library based on GotoBLAS2 1.13 BSD version.

verso. Lean documentation authoring tool

PufferDrive. Continued development of pufferdrive

lean4web. Lean web editor

open-r1. Fully open reproduction of DeepSeek-R1

boo. A GNU screen style terminal multiplexer built on libghostty.

burr. Build applications that make decisions (chatbots, agents, simulations, etc...). Monitor, trace, persist, and execute on your own infrastructure.

PyOMP. OpenMP for Python through Numba

velvet. An auto-active verifier embedded into Lean

veil. A verifier for automated and interactive proofs about transition systems.

TorchLean. TorchLean is the first unified Lean 4 framework for neural-network specification, execution, and verification.

Pantograph. (Mirror) A Machine-to-Machine Interaction System for Lean 4

LLVM_RISCV. Lean

forest. VSCode extension for parallel feature development using git worktrees.

defending-code-reference-harness. Skills for threat modeling, scanning, triage, patching, plus an autonomous scanning harness you can /customize

openvscode-server. Run upstream VS Code on a remote machine with access through a modern web browser from any device, anywhere.

openwhispr. Voice-to-text dictation app with local (Nvidia Parakeet/Whisper) and cloud models (BYOK). Privacy-first and available cross-platform.

pierre. pierre’s open source code

stable-worldmodel. A platform for reproducible world model research and evaluation

lemmascript-kit. starter kit for LemmaScript

oh-my-pi. ⌥ AI Coding agent for the terminal — hash-anchored edits, optimized tool harness, LSP, Python, browser, subagents, and more

netron. Visualizer for neural network, deep learning and machine learning models

TorchSpec. A PyTorch native library for training speculative decoding models

EAGLE. Official Implementation of EAGLE-1 (ICML'24), EAGLE-2 (EMNLP'24), and EAGLE-3 (NeurIPS'25).

tueplots. Figure sizes, font sizes, fonts, and more configurations at minimal overhead. Fix your journal papers, conference proceedings, and other scientific publications.

tectonic. A modernized, complete, self-contained TeX/LaTeX engine, powered by XeTeX and TeXLive.

Tadpole. Flexible scientific foundation models trained with synthetic online data generation

silk. Fast stackful fibers with a NUMA-aware work-stealing scheduler

ax. Google's open source distributed agent runtime

archsem. Rocq framework to define the semantics of CPU architectures

opencode-lemmascript. LemmaScript case study of OpenCode: The open source coding agent.

coder. Secure environments for developers and their agents

volcano. A Cloud Native Batch System (Project under CNCF)

xsimd. C++ wrappers for SIMD intrinsics and parallelized, optimized mathematical functions (SSE, AVX, AVX512, NEON, SVE, WebAssembly, VSX, RISC-V))

highway. Performance-portable, length-agnostic SIMD with runtime dispatch

Vc. SIMD Vector Classes for C++

cv. Curriculum Vitae of Luis Wirth

SciLean. Scientific computing in Lean 4

cslib. The Lean Computer Science Library (CSLib)

skills. Skills for AIs using the Lean programming language and theorem prover — proofs, toolchain setup, bisection, and more

DeepSeek-Prover-V1.5. Python

LeanCopilot. LLMs as Copilots for Theorem Proving in Lean

pytorch-hccl-tests. Torch distributed benchmarks based on PyTorch/Ascend adapter.

claude-for-legal. A suite of plugins for legal workflows

click. Python composable command line interface toolkit

formula_gan. Unsupervised Training Data Generation of Handwritten Formulas using Generative Adversarial Networks with Self-Attention

HMEG. [CVPR'24] Handwritten Mathematical Expressions Generation (HMEG)

pi-llamacpp. An experimental pi extension that runs and manages qwen with llama.cpp

ds4. DeepSeek 4 Flash and PRO local inference engine for Metal, CUDA and ROCm

pi-ds4. Run deepseek4 locally on metal right from within Pi

Uni-MuMER. [NeurIPS'25 Spotlight🔥]Official implementation of Uni-MuMER: Unified Multi-Task Fine-Tuning of Vision-Language Model for Handwritten Mathematical Expression Recognition

dafny-replay-bench-baselines. Baselines for dafny-replay benchmark

dynamicLRP. Model-agnostic and differentiable LRP in PyTorch that runs on the autograd graph at the operation level. Implements the latest work in LRP literature and extends the premise of LRP to fundamental PyTorch tensor ops.

bocpy. Behavior-Oriented Concurrency in Python

hunk. Review-first terminal diff viewer for agentic coders

cc-lemma. Refreshing automation for inductive equational proofs using e-graphs

flash-linear-attention. 🚀 Efficient implementations for emerging model architectures

caveman. 🪨 why use many token when few token do trick — Claude Code skill that cuts 65% of tokens by talking like caveman

pykaleidoscope. Implementation of the LLVM tutorial in Python using llvmlite

text-to-cad. A collection of agent skills for CAD, robotics and hardware design

practical-fm. A gently curated list of companies using verification formal methods in industry

xn-ptts. Pocket-TTS inference using xn

hn-virality-predictor. Python

asv. Airspeed Velocity: A simple Python benchmarking tool with web-based reporting

gemini-cam. A Beamerposter template with University of Cambridge logo and colors. It is forked from Gemini.

whatcable. macOS menu bar app that tells you, in plain English, what each USB-C cable plugged into your Mac can actually do

lib0xc. Safe(ish) C programming library

elan. The Lean version manager

equality-game-lemmascript. LemmaScript case study of a game where the goal is to make two sides equal

github-star-checker-lemmascript. LemmaScript case study of a CLI to check star counts on GitHub

clear-split-lemmascript. LemmaScript case study

colorwheel-lemmascript. LemmaScript case study

pyptx. A Python DSL to write Nvidia PTX for Hopper and Blackwell in JAX and PyTorch

guardians. an implementation of the ideas in Erik Meijer's "Guardians of the Agents: Formal Verification of AI Workflows" (Communications of the ACM, January 2026)

vericoding-benchmark. Lean

deductive-vericoding. Tactics-driven vericoding

devlogs. Developer blog for PyTorch

parameter-golf. Train the smallest LM you can that fits in 16MB. Best model wins!

gccrs. GCC Front-End for Rust

mise. dev tools, env vars, task runner

TorchCode. 🔥 LeetCode for PyTorch — practice implementing softmax, attention, GPT-2 and more from scratch with instant auto-grading. Jupyter-based, self-hosted or try online.

lodepng-afl-fuzz-differential. Differential fuzzing harness for lodepng-rust

ml-compiler-opt. Infrastructure for Machine Learning Guided Optimization (MLGO) in LLVM.

lambench. TypeScript

constrained-decoding. A guide to structured generation using constrained decoding

ghstack. Submit stacked diffs to GitHub on the command line

hypothesis. The property-based testing library for Python

apytypes. APyTypes - Algorithmic data types for Python

cal.diy. Scheduling infrastructure for absolutely everyone.

Reactant.jl. Optimize Julia Functions With MLIR and XLA for High-Performance Execution on CPU, GPU, TPU and more.

xdsl-autotuning-paper-experiments. MLIR

deltakit. Riverlane's quantum error correction SDK

jai. Jail your AI agent

pychop. A Python package for simulating low precision arithmetic in scientific computing and machine learning

liburing. Library providing helpers for the Linux kernel io_uring support

foundationdb. FoundationDB - the open source, distributed, transactional key-value store

eudsl. Embedded Universal DSL: a good DSL for us, by us

smolvm. Portable, lightweight, self-contained virtual machines.

souper. A superoptimizer for LLVM IR

alive2. Automatic verification of LLVM optimizations

diffrax. Numerical differential equation solvers in JAX. Autodifferentiable and GPU-capable. https://docs.kidger.site/diffrax/

collab-todo-lemmascript. LemmaScript case study

LemmaScript. verification toolchain for TypeScript (Tech Preview)

ovie. Official implementation and models for OVIE (One View Is Enough! Monocular Training for In-the-Wild Novel View Generation)

lean4. Lean 4 programming language and theorem prover

ParseBench. ParseBench - A Document Parsing Benchmark for AI Agents

cli. Google Workspace CLI — one command-line tool for Drive, Gmail, Calendar, Sheets, Docs, Chat, Admin, and more. Dynamically built from Google Discovery Service. Includes AI agent skills.

Handy. A free, open source, and extensible speech-to-text application that works completely offline.

nodejs-polars. nodejs front-end of polars

wiiMac. A Mac OS X bootloader for the Nintendo Wii.

tailslayer. Library for reducing tail latency in RAM reads

boltz-cp. Context Parallelism Code for Boltz-2

cuEquivariance. cuEquivariance is a math library that is a collective of low-level primitives and tensor ops to accelerate widely-used models, like DiffDock, MACE, Allegro and NEQUIP, based on equivariant neural networks. Also includes kernels for accelerated structure prediction.

alphafold. Open source code for AlphaFold 2.

claude-code. Claude Code is an agentic coding tool that lives in your terminal, understands your codebase, and helps you code faster by executing routine tasks, explaining complex code, and handling git workflows - all through natural language commands.

guppylm. A ~9M parameter LLM that talks like a small fish.

bsc. Bluespec Compiler (BSC)

github-historical-uptime. Plotting the historical uptime of GitHub

learn-coding-agent. Research on Coding Agents

claw-code. An agent-managed museum exhibit, built in Rust with Gajae-Code / LazyCodex — developed and maintained with no human intervention.

fernanda.ink. a collection of thoughts on building with and for AI

superset. Code Editor for the AI Agents Era - Run an army of Claude Code, Codex, etc. on your machine

Autograd.hs. Automatic differentiation engine in Haskell.

markit. 🖍️ Convert anything to markdown. Mark it.

slides. Terminal based presentation tool

benchmarks. Open benchmarks for evaluating search APIs

LiteRT. LiteRT, successor to TensorFlow Lite. is Google's On-device framework for high-performance ML & GenAI deployment on edge platforms, via efficient conversion, runtime, and optimization

RustTraining. Beginner, advanced, expert level Rust training material

worldmonitor. Real-time global intelligence dashboard. AI-powered news aggregation, geopolitical monitoring, and infrastructure tracking in a unified situational awareness interface

puppeteer. JavaScript API for Chrome and Firefox

Engram. Conditional Memory via Scalable Lookup: A New Axis of Sparsity for Large Language Models

SuperCombinators. Parser combinator framework written in Swift

superlopuh.github.io. My website.

lz-string. A reimplementation of the lz-string node package.

array-api-compat. Compatibility layer for common array libraries to support the Array API

vickrey. Lean 4 formalization of strategy-proofness for the Vickrey (second-price sealed-bid) auction

alphadev. Python

oryx. Oryx is a library for probabilistic programming and deep learning built on top of Jax.

models.dev. An open-source database of AI models.

doesjitgobrrr. A performance dashboard tracking CPython's JIT vs. Interpreter benchmarks.

cpython. The Python programming language

ein. Points for Free: Embedding Pointful Array Programming in Python

openfold. Trainable, memory-efficient, and GPU-friendly PyTorch reproduction of AlphaFold 2

gpn. Genomic Pre-trained Network