focusbulb
I wired a smart bulb to turn red when I'm in a meeting
Polls my calendar and flips a cheap smart bulb red during calls so my family knows not to walk in. Twenty lines of Python and a webhook. Best ROI of the year.
1
Whisperdesk
I made offline meeting transcription that never leaves your laptop
Runs Whisper locally and drops a clean, speaker-labeled transcript into a plain markdown file. No cloud, no upload, it works on a plane and it works when the meeting was about something you would rather not hand to a third party. I built it after reading one too many transcription-tool privacy policies that quietly reserved the right to train on my audio. Speaker diarization without the cloud was the fun part: a small local embedding model gives rough labels that are good enough to skim a meeting you missed. It still trips on heavy crosstalk and long recordings take a while, but it has become the thing I reach for every week.
3
Quantize
I got small LLMs running nicely on a Raspberry Pi
A wrapper plus web UI around llama.cpp tuned for single board computers like the Raspberry Pi. It streams tokens over websockets and stays under the Pi's tiny memory budget using Q4 quantization and mmap, so a 3B model actually fits in 2GB. Tokens per second are slow, but for a home assistant that mostly answers short questions it is real-time enough to feel responsive. I wanted something that ran entirely on hardware I own: no API keys, no usage caps, and no sending my house's voice commands off to someone's cloud. Setup is a single script that auto-detects the board. Local Whisper for voice input is the next milestone.
1