alvdansen/animating-on-twos
Animating on Twos: Training Keyframe-Animation Adapters on a Pretrained Video Model — Alvdansen Labs, August 2026
01 / MACHINE INTELLIGENCE
Models you can run. Agents you can inspect. Tools you can build on.
← All topicsAnimating on Twos: Training Keyframe-Animation Adapters on a Pretrained Video Model — Alvdansen Labs, August 2026
A curated list of community plugins for bb, the agent IDE — the ecosystem's missing directory.
Your agent seeks what search can't find. A self-hosted perception MCP server that transcribes speech, reads behind logins, sees images and video frames, crosses languages, and remembers.
An open-source workspace for coding agents, running on your own machine. More work. Less AI waste. Same subscriptions.
The review app for code your agent writes. Native Rust/GPUI desktop app with real Git workflows.
Foundation Models for Structured Data.
Interactive data exploration that is extremely fast and agent-native
BIMwright — AI-native Revit MCP server (Revit 2022–2027)
Open the project to explore the idea and the code.
Open the project to explore the idea and the code.
A curated list of AI-powered pentesting tools, frameworks, MCP servers, and resources for autonomous cybersecurity operations
Harness-Zero: Harness Distillation via Agent-as-Harness
co/core is a place where people share the compute they already own to run AI for each other, instead of renting from a handful of giant providers
Production-grade screenplay studio: converts bullet points & images into strict single-line, timecoded video prompts for MiniMax H3, Maestro, Kling & Runway. Local LM Studio vision, floorplan analysis, DE/EN GUI. Zero line breaks guaranteed.
An Elixir/OTP AI agent runtime: define agents, connect any OpenAI-compatible model, and run a tool-calling loop. CLI, HTTP API, web dashboard, Telegram, WhatsApp.
The open archive of how AI agents actually work. A versioned, verifiable archive of system prompts, developer instructions, tool schemas and agent harnesses powering today's AI products. Captured with OrcaReplay.
A experimental and feature-rich inference engine built on top of llama.cpp
RSIGame: Autonomous agentic game development with recursive self-improvement
A lightweight, native Swift 6 bridge integrating TypeSafe AI's Jev System One decision model into Apple's Foundation Models framework. - peterfriese/system-one-foundation-models
A lightweight, native Swift 6 bridge integrating TypeSafe AI's Jev System One decision model into Apple's Foundation Models framework.
Open decisions model with Jev's API: typed answers (yes/no, choice, score, multi) with probabilities from forward passes, no generation. Qwen3.5-4B + LoRA, one GPU.
Kyojin: the Yamz inference engine for AMD Strix Halo (ROCm, gfx1151), built on ExLlamaV3. Runs 300B-class MoE models on one 128 GB mini PC.
Open-source CLI & TUI to port AI coding sessions across Claude Code, Codex, and OpenCode.
The durable runtime for AI agents: event-sourced sessions, approvals that persist across processes, exact resume after a crash. A 2,200-line TypeScript kernel, and kiso-code, the coding agent built on it.