All of my long-form thoughts on programming, cooking, and more, collected in chronological order.
I have a Model Y on order. I also have an RTX 3090 and a habit of building tools out of frustration. The result: 8 dependency-ordered Tesla repos, 6 of them live on PyPI, 611 passing tests, and a workspace that any owner can run without a vehicle in hand.
42 chapters. 60 seconds each. Local GPU. The Bible is public domain, AI video gen is finally usable, and routing all the LLM work through my Claude Code subscription means the per-render cost is $0.
Snap a photo of your plate, let a vision model name what's on it, and back the macros from free USDA data. Offline-first, ad-free, mine. Built because I got tired of MyFitnessPal's UX regressions and paywalls.
Most prayer apps are dressed-up journals. The differentiator I wanted was felt presence — when you post a request, real people pray for it. When you open the app, you can carry a stranger's burden for two minutes.
Built on the MacArthur-Bates Communicative Development Inventories — the actual instrument speech-language pathologists use — instead of yet another habit-tracker UI. Flutter, Isar, COPPA-aware, bilingual-aware.
Faster-whisper transcribes the videos. pyannote handles diarization. BGE embeddings land in ChromaDB. LM Studio answers questions with timestamped citations. Ten phases shipped end-to-end and nothing leaves my desktop.
Most bots run a single script forever. I wanted one with goals — combat to 70, wealth to 1M GP, 200+ QP — that picks its own next action based on game state, inventory, and how close it is to each target. YOLOv8 for vision, an LLM for strategy, a state machine for everything else.
Used eye of the round for the meat, hand sliced from whole Costco cut. Used dragons milk dark stout for the beer in marinade, and everything else listed in recipe which we doubled.