Projects Recently Starred by Experts

Starred 5 hours ago by

Peter Norvig(Author of "Artificial Intelligence: A Modern Approach"; Research Director at Google).

reasoning-bank by google-research

31.1%

256

Agent self-evolution via reasoning memory and scaling

Enhances agent performance by learning from past successes and failures using stored reasoning.
Enables memory-aware test-time scaling, treating experience as a new dimension for agent systems.
Integrates GPT, Gemini, and Claude LLMs for web browsing and software engineering benchmarks.
Utilizes LLM-as-a-judge for robust evaluation of agent trajectories.

Created 2 months ago

Updated 1 week ago

Starred 6 hours ago by

Didier Lopes(Founder of OpenBB). Also starred by

Junyang Lin(Core Maintainer at Alibaba Qwen),

Li Jiang(Coauthor of AutoGen; Engineer at Microsoft),

Chip Huyen(Author of "AI Engineering", "Designing Machine Learning Systems"), and

1 more.

TradingAgents by TauricResearch

10.5%

58k

Multi-agent LLM framework for financial trading (research paper)

LLM-powered agents for market analysis and trading decisions.
Designed for AAAI 2025 MARW Workshop.

Created 1 year ago

Updated 5 days ago

Starred 14 hours ago by

Wes McKinney(Author of Pandas). Also starred by

Elie Bursztein(Cybersecurity Lead at Google DeepMind),

Theo Browne(Founder of Ping.gg),

Jeff Hammerbacher(Cofounder of Cloudera), and

8 more.

symphony by openai

31.3%

20k

Autonomous agents for managing software development workflows

Automates project tasks by spawning autonomous agents that handle work and provide proof of completion.
Integrates with task boards to monitor work and orchestrates agent execution for implementation runs.
Shifts focus from supervising coding agents to managing work at a higher level.
Agents deliver comprehensive proof of work: CI status, PR feedback, complexity analysis, and walkthrough videos.

Created 2 months ago

Updated 3 days ago

Starred 14 hours ago by

Shyamal Anadkat(Research Scientist at OpenAI). Also starred by

Chip Huyen(Author of "AI Engineering", "Designing Machine Learning Systems"),

Didier Lopes(Founder of OpenBB),

Elie Bursztein(Cybersecurity Lead at Google DeepMind), and

4 more.

qlib by microsoft

1.0%

42k

AI platform for quantitative investment research and production

Supports supervised learning, market dynamics modeling, and reinforcement learning.
Includes data processing, model training, back-testing, and portfolio optimization.
Features LLM-driven autonomous agents for factor mining and model optimization.
Provides a model zoo with GBDT, LSTM, Transformer, and other models.

Created 5 years ago

Updated 1 week ago

Starred 17 hours ago by

Georgios Konstantopoulos(CTO, General Partner at Paradigm). Also starred by

Pawel Garbacki(Cofounder of Fireworks AI),

Luis Capelo(Cofounder of Lightning AI),

Solomon Hykes(Cofounder of Docker, Dagger), and

2 more.

zml by zml

0.7%

AI inference stack for production

Runs LLaMA2 sharded across NVIDIA, AMD GPUs & Google Cloud TPU, activations passed over VPN.
Uses Zig, MLIR, Bazel for high-performance, cross-compilation on Mac.
Supports CUDA, RoCM, TPU, AWS Trainium/Inferentia 2 accelerators.
Includes MNIST example and Meta Llama 3.1/3.2 models.

Created 1 year ago

Updated 16 hours ago

Starred 19 hours ago by

John Resig(Author of jQuery; Chief Software Architect at Khan Academy).

agent-sprite-forge by 0x0funky

454.9%

2D game asset generation from prompts

Creates sprite sheets, animations, and FX for characters, spells, and creatures using natural language.
Generates layered maps with collision data, props, and zones for playable game environments.
Agentic workflow integrates image generation and deterministic local post-processing for clean asset pipelines.
Exports transparent PNG frames, animated GIFs, and game-ready assets via a unified agent.

Created 1 week ago

Updated 16 hours ago

Starred 1 day ago by

Matteo Collina(Cofounder of Platformatic). Also starred by

Travis Fischer(Founder of Agentic).

markit by Michaelliv

2.1%

Universal document and media converter with AI integration

Converts diverse formats (PDF, DOCX, images, URLs) to Markdown via CLI or SDK.
Leverages LLMs for image description and audio transcription from multimodal inputs.
Extensible architecture with pluggable converters and custom AI provider support.
Enables structured data extraction and AI-driven content analysis.

Created 1 month ago

Updated 2 weeks ago

Starred 1 day ago by

Vincent Weisser(Cofounder of Prime Intellect).

predict-rlm by Trampoline-AI

69.0%

261

Production Recursive Language Model runtime

Enables LMs to call sub-LMs with DSPy signatures for complex, interpretable control flow.
Supports multimodal inputs (images, docs, audio, video) and structured data I/O via typed files.
Avoids context rot and scales performance directly with base model improvements, unlike traditional harnesses.
Integrates with DSPy, Pydantic, and supports async tool calling within a WASM sandbox.

Created 1 month ago

Updated 14 hours ago

Starred 1 day ago by

Didier Lopes(Founder of OpenBB). Also starred by

Jonathan Ragan-Kelley(Professor at MIT),

Travis Fischer(Founder of Agentic),

Georgi Gerganov(Author of llama.cpp, whisper.cpp), and

3 more.

Handy by cjpais

1.5%

21k

Offline, extensible speech-to-text application

Transcribe speech to any text field locally, prioritizing privacy and offline use.
Leverages Whisper for transcription and VAD for silence filtering, with GPU acceleration.
Built with Tauri (Rust/React/TypeScript) for cross-platform desktop compatibility.
Offers a forkable foundation for community development and customization.

Created 1 year ago

Updated 1 day ago

Starred 1 day ago by

Luis Capelo(Cofounder of Lightning AI). Also starred by

Jiaming Song(Chief Scientist at Luma AI) and

Jeff Hammerbacher(Cofounder of Cloudera).

Automodel by NVIDIA-NeMo

5.9%

477

PyTorch-native SPMD library for LLM/VLM training

Scales LLM/VLM training and finetuning using PyTorch Distributed and SPMD.
Offers seamless Hugging Face integration for LLMs and VLMs, enabling day-zero support.
Utilizes DTensor, FSDP2, and pipeline parallelism for efficient distributed training.
Features FP8 precision and custom kernels for optimized training performance.

Created 11 months ago

Updated 8 hours ago

...