Carlos Justiniano

iac lets Claude Code, Codex, ChatGPT and local models hand each other work as JSON comments on a private GitHub repo. No server to host.

explain covers a concept in 250 words, or as a diagrammed HTML page when the shape is the point.

council puts one hard question to several independent members (Claude by default, with Codex, Ollama or OpenRouter optional) and reconciles the answers. It won't manufacture consensus, and it tells you the cost multiplier before it spends.

bookcraft writes a book from a one-line description, binds it into a PDF and EPUB with figures and an index, revises chapters in place, and checks each paraphrased claim against the source it cites.

I've packaged my Claude Code skills into a public plugin marketplace. 5 plugins, 36 skills, MIT licensed. claude plugin marketplace add cjus/cjus-skills What's inside 🧵

Working on nextgen chess software - more to come. #chess #rustlang

Learn how to build a cross-platform chess engine in Rust. 360+ page guide included. github.com/cjus/chessengine

In 2004 Colin Frayn and I set a Guinness World Record with ChessBrain, the largest networked chess computer. Some ideas lingered, so I created cb2026 (ChessBrain 2026) to explore them: a chess engine built from scratch in Rust, ~3379 CCRL Blitz. Open source: github.com/cjus/chessengine #rustlang #chess

Just published bevy-game-recorder 🎥 Records your Bevy game's window and audio to an MP4 on macOS. Built on ScreenCaptureKit: no FFmpeg, no GPU readback, and the title bar stays out of the video. cargo add bevy-game-recorder github.com/cjus/bevy-game-re… #bevy #rustlang

Yesterday I just open-sourced audio-chat-box: give your AI project a voice. It speaks and listens, hands-free and fully offline. Nothing leaves your machine. github.com/cjus/audio-chat-b… Try it in one command, nothing to build (models included): docker run --rm --platform linux/amd64 -p 127.0.0.1:8080:8080 pnxstudios/audio-chat-box-demo Then open localhost:8080, press Record and talk. How it works: 🔊 Kokoro reads text aloud, in 54 voices 👂 NVIDIA's Parakeet turns speech into text 🎙️ Silero VAD listens as you talk; Parakeet transcribes each thing you say when you pause Your program starts it and exchanges JSON messages over stdin/stdout, one per line. Python, Node, Go, Rust, Electron: anything that can start a process can use it. It never touches the mic or speakers; your app records and plays. Hotwords: tell it the words your product expects and it favours them where the audio is ambiguous. In a chess app, "knight" beats "night". Built natively on an M1 Max, both models load in about a second and a sentence is spoken in under one. (The Docker image is amd64; on Apple Silicon it runs under emulation and is slower.) It began as the voice of Solrac (solrac-chess.com), my AI chess coach. I took out everything chess-specific so any project can use it. Today: using it in your own app means building from source (Rust). English transcription, one listen at a time. Code MIT / Apache-2.0; the program as a whole GPL-3.0, since it includes eSpeak NG.

I just open-sourced audio-chat-box: give your AI project a voice. It speaks, listens and understands, hands-free and fully offline. Nothing leaves your machine. github.com/cjus/audio-chat-b… How it works: 🔊 Kokoro reads text aloud, in 54 voices 👂 NVIDIA's Parakeet turns speech into text 🎙️ Silero VAD listens as you talk, and Parakeet transcribes each thing you say when you pause Your program starts it and exchanges JSON messages over stdin/stdout, one per line. Python, Node, Go, Rust, Electron: anything that can start a process can use it. It never touches the mic or speakers; your app records and plays, so it fits however your app already handles audio. Hotwords: tell it the words your product expects, and it favours them where the audio is ambiguous. In a chess app, "knight" beats "night". In yours, your product's names. On an M1 Max both models load in about a second, and a sentence is spoken in under one. The models (~835 MB) download once and are checked against SHA-256s built into the program. Run it with --offline and it never touches the network. It began as the voice of Solrac (solrac-chess.com), my AI chess coach, where its listening was tuned on moves said at speed-chess pace. I took out everything chess-specific so any project can use it. Today: build from source (Rust), English transcription, one listen at a time. Code is MIT / Apache-2.0; the program as a whole is GPL-3.0, since it includes eSpeak NG. Your app runs it as a separate program and doesn't link it. Feedback and PRs welcome, especially if you're adding voice to an agent or assistant. ⭐ if it's useful. #rustlang #OpenSource

3D Chess - Rust, Bevy

I spent two years building an open-source and commercial product, in Rust, with my friends at @514hq. Now I’m deep in agentic offensive security. So the Rust I worked hard to learn doesn’t fade, I’m using it on some long-standing computer chess ideas. ♟️

Been working on an interesting project for my fellow #qigong #taichi #kungfu #taoist practitioners and instructors. qigong-with-carlos.com

Tony Robbins takes center stage to a standing ovation. #SlopesSummit23 – at Salt Lake City, UT

Jersey boy Gary Vee discusses his approach to leveraging attention in order to capitalize on short-term trading opportunities using social media. #SlopesSummit23

3D rendering of a Chinese training hall for my up coming Taoist inspired mobile app.