A curated list of cool original open-source AI projects I've personally interacted with — starred, forked, mentioned, linked, or commented on. No personal forks listed, only the upstream originals.
Focus: agents, coding tools, inference, VLMs, biomed, and foundational generative AI.
- My Own Original Projects
- Original Projects I've Starred
- Original Projects I've Forked (Upstream Repos)
- Original Projects & Models I've Mentioned or Linked on X
- Hugging Face Models & Collections Mentioned
- Foundational / Historical Projects Mentioned
- Notes
- Contributing
- License
- SIPS Self Improvement Plugin System — Free open-source OpenCode/Codex system with various live-service hooks, memory, etc. Very WIP.
- StorageScope — Free open-source macOS disk-space analyzer and storage cleaner (Swift/SwiftUI). Duplicate-file review, large-folder scanning, local-first cleanup planning. Interaction: Created, maintained, and released multiple versions (v0.3.0 → v0.6.0+); promoted on X. Developed with assistance from GLM 5.2 and Claude.
- Open-Software-Network/OS-June — Open-source desktop app that turns spoken work into useful work: reliable local audio recording, editable meeting notes, cleaned dictation pasted into any app, and a local agent for files, research, drafts, and routines. Venice-powered privacy Rampart Redaction Model mentioned for June WIP. Interaction: Starred.
- StarTrail-org/PixelRAG — Pixel-native AI search — the end of traditional web parsing and the beginning of scalable pixel-based search. Interaction: Starred.
- Panniantong/Agent-Reach — Give your AI agent eyes to see the entire internet: a CLI to read/search Twitter, Reddit, YouTube, GitHub, etc. (zero API fees). Interaction: Starred.
- openai/codex — Lightweight coding agent that runs in your terminal. Interaction: Starred.
- browser-use/browser-harness — Self-healing harness that enables LLMs to complete any browser-based task. Interaction: Starred.
- Blaizzy/mlx-vlm — Inference and fine-tuning of Vision-Language Models (VLMs) on Apple Silicon Macs using MLX. Interaction: Starred.
- ultraworkers/claw-code — Autonomous, agent-managed AI systems and code generation (highly active autonomous-agent project). Interaction: Starred.
- poetiq-ai/poetiq-arc-agi-solver — Reproduction of record-breaking submissions to the ARC-AGI-1 and ARC-AGI-2 benchmarks. Interaction: Starred.
- snap-stanford/Biomni — General-purpose biomedical AI agent. Interaction: Starred.
- Agenta-AI/agenta — Open-source LLMOps platform: prompt playground, management, evaluation, and observability. Interaction: Starred.
- Adriankhl/godot-llm — Integrate LLMs directly into the Godot game engine. Interaction: Starred.
These are the original repositories I forked:
- Noumena-Network/code — Coding TUI optimized for AI inference stacks.
Interaction: Forked (as
code). - StarTrail-org/PixelRAG — Pixel-native AI search. Interaction: Forked (also starred).
- NousResearch/hermes-agent — Self-improving AI agent that grows with the user. Interaction: Forked.
- Panniantong/Agent-Reach — AI agent internet search CLI. Interaction: Forked (also starred).
- andrewmccalip/kickbacks.ai — Plugin for Claude Code, Codex, and Hermes Agent (clickable ads in real developer sessions). Interaction: Forked + commented on GitHub issues.
- zhaoxuya520/reverse-skill — AI-powered reverse-engineering / penetration-testing skill router with a self-evolving knowledge base (supports Claude Code). Interaction: Forked.
- Gitlawb/openclaude — Flexible, runs-anywhere Claude-related open tool. Interaction: Forked.
- NVIDIA-BioNeMo/bionemo-agent-toolkit — Agent toolkit for chemistry, proteins, and biomolecular AI research. Interaction: Posted about on X ("Pretty cool").
- AliaksandrSiarohin/first-order-model — Classic open-source model for image animation and motion transfer. Interaction: Shared link on X.
- openai/plugins (eval plugin and related tools) Interaction: Shared link on X.
- VQGAN + CLIP ecosystem (foundational generative AI) — Example repo: justinjohn0306/VQGAN-CLIP and related implementations. Interaction: Referenced in the context of early AI experimentation.
These are open models/collections I directly linked or discussed:
- deepreinforce-ai/ornith-10 (Hugging Face collection) — High-quality models distilled from Gemma 4 + Qwen. Interaction: Shared link and discussed performance on X.
- Gemma-4 variants
- google/gemma-4-E2B-it-qat-mobile-transformers
- WebGPU inference: webml-community/Gemma-4-WebGPU Interaction: Shared links and discussed optimizations / Fable performance.
- prefeitura-rio/Rio-3.5-Open-397B — Large open 397B model released by Rio's government. Interaction: Shared and discussed on X.
- Qwen models (Alibaba open-weight releases) Interaction: Commented on open-source strategy and rapid community progress.
- EleutherAI / EleutherLabs ecosystem (GPT-J, Pythia, and early open LLM work) Interaction: Referenced personal history starting with their LLM testing and VQGAN+CLIP experiments.
- Only original upstream repositories are listed.
- Interaction types are noted for context (starred, forked, posted about, etc.).
- Strong emphasis on agentic systems, local inference, coding agents, and practical open-source tools.
- Open-source is the way — these projects represent real acceleration and accessibility in AI.
Suggestions or additions are welcome! Open an issue or pull request, or just tag me on X.
Released under the MIT License. Curated by RasputinKaiser.