A small Bitcoin-like chain whose canonical ledger is my README. Proof of work is subset-sum, not hashing.
A small, educational vLLM implementation for exploring LLM inference, scheduling, KV-cache management, and GPU execution.
Agent Skills for small-molecule and protein therapeutics: cheminformatics, molecular ML, docking and dynamics, protein design platforms, and target-discovery knowledge graphs.
A searchable Argolink guide for DeepSeek V4 Pro: the chat-compatible chat contract, reasoning controls, pricing units, operational errors, and a small offline content tree.
Self-improvement loop for Hermes Agent: reads its own trajectory, proposes and applies the smallest skill/memory edit (create/patch), journaled with rollback. /refine command, refine_run tool, on_session_end hook.
Sparse-upcycle a dense Gemma-3-270m into a Mixture-of-Experts (gemma3moe) and serve it on llama.cpp (CPU). Upcycle, train, inspect routing, GGUF.
Small, fast, free: an open recipe for a sub-100M TTS. Code, data, weights. The PoC trained to a working model in a few hours on one RTX 3090, and inspired a wave of open-weights TTS releases.
A small plugin for Claude Code and Codex to optimize PyTorch workloads.
Community Core ML runtime, Swift SDK, and iOS/macOS samples for Irodori TTS v4.1 Small MF.
Meet KaLM-Jev β your local, Jev-style judgment engine, available in Nano, Small, and Large.
Small System One decision models (17Mβ400M) for Yes/No, Choice, Score, and reranking. Training toolkit, Python inference, and ONNX support for Node.js and browsers.
A small, auditable cost estimator for long-running Claude Opus 5.5 calls through Argolink.
Local voice control for Home Assistant on a Raspberry Pi 5: streaming speech-to-text (EdgeSTT) and a small non-generative decision engine (EdgeDecision). 9 languages, no cloud, no LLM.
End-to-end spoken language understanding toolkit: a small trainable neural audio codec with RVQ, joint intent-and-slot models over speech features, streaming partial intents, a speech-LLM instruction data pipeline, and turn-level evaluation with reproducible reports. NumPy core with a CPU-only torch extra; synthetic data only, fully offline.
WILRO-MoE: a small VLA for LIBERO, with golden-ticket noise search and deterministic noise cycles
εΌε ε°η―εδ½
A way of running a project with coding agents. A team of roles, a loop with a stop in it, and rules your agent cannot ignore. Text files and small Python programs, no dependencies.
Packs text files into small images that an AI model reads for about half the input tokens. Includes a Claude Code plugin, a right-click tool and an HTML app.
A tiny jev-like model that answers Choice, Score and Noul questions in one forward pass and returns calibrated probabilities. MLX or PyTorch, fully offline, System One compatible.
A small, explainable AI workflow for reviewing Chinese DOCX contracts.
MicroK8s is a small, fast, single-package Kubernetes for datacenters and the edge.
ζ ι欑ζ°ηbanana proοΌιε账ε·ζ± οΌζ―ζθ΄θ½½εθ‘‘γATθͺε¨ε·ζ°γηΌεηη₯γ代ηηγQδΊ€ζ΅ηΎ€1073237297
Small offensive network toolkit for Raspberry Pi (+ Waveshare 1.44β³ LCD HAT) and Cardputer Zero inspired by pager and sharkjack fonctionnalities. For redteam and educational purposes only.
Active Learning for Text Classification in Python
A small Python library and CLI for working with Cloudflare Turnstile on your own pages - read the sitekey, build the token, verify the result.
A small Python library and CLI for working with Cloudflare Turnstile on your own pages - read the sitekey, build the token, verify the result.
AgentTouch: a small face on your desk for your AI coding agents (Waveshare ESP32-S3-Touch-AMOLED-2.16 + macOS host)