I build practical AI systems and developer tools, ship small experiments, learn in public, and keep software a little playful.

All my agent skills for Codex and Claude Code.

• Shipyard - Coordinate Codex and Claude with scoped tests, model routing, CI acceptance, and safe branch cleanup.
• bp - Open apps, sites and files on macOS or Linux from saved nicknames, run tasks across apps, and learn new shortcuts as you work.
• remote-browser - Run a cloud browser on your Cloudflare account for JavaScript pages and manual logins.
• webpage-to-pdf - Save a live webpage as a PDF, pixel-exact or with searchable text.
• go-workspace-skill - Sync, build, and test a multi-repo Go workspace from one config file.
• stitch - Generate and edit Google Stitch designs and turn them into app code.
• runlens - Find failures, flaky tools, and latency outliers in agent JSONL traces.
• ragcheck - Score retrieval runs with precision, recall, MRR, MAP, and nDCG@k, and check answers by word overlap.
• promptdeck - Render prompt templates, expand variant matrices, and rank scored variants.
• datasetlint - Audit JSONL datasets for duplicates, train/eval leakage, and label conflicts.

• gauth2apple - A tool that splits a Google Authenticator export QR code into individual otpauth:// QR codes for Apple Passwords.
• workspace - A Makefile-driven workspace for syncing and managing my project repos.
• homebrew-tap - A Homebrew tap for casks maintained and shipped under my namespace.
• go-chrome-ai - A cross-platform Chrome profile patcher written in Go, with CLI and GUI modes, that enables Chrome AI features such as Ask Gemini and can block silent Gemini Nano downloads.
• dotfiles - A chezmoi-managed personal dotfiles repo for Neovim, Starship, and reproducible local setup.

• kitten-asr-go - A Go implementation of KittenML's lightweight ONNX ASR models for speech-to-text.
• kitten-tts-go - A Go implementation of KittenTTS, a lightweight ONNX text-to-speech engine.
• delta - An end-to-end, TensorFlow-based deep learning platform for training and deploying natural language and speech models.
• dstc9 - DSTC9 dialog-system challenge code, experiments, and related research work.

• Turning day-to-day automation patterns into installable agent skills.
• Exploring workflows across Go, Python, and LLM applications.
• Shipping products that run on Docker and Kubernetes.

Make it useful. Keep it light. Ship it.