Pattern is three things. One is a (work-in-progress) service, an AI personal assistant for neurodivergent people, architected as a constellation of specialized LLM (Large Language Model, for those unfamiliar, the thing that powers ChatGPT) agents. Another is a framework, my own take on memory-augmented LLM agents, written entirely in Rust. A longer-form overview blog post I wrote back when I first brought Pattern to Bluesky is available here. The third is, well, @pattern.atproto.systems and you can go talk to Them.

Pattern was developed in something of a fit of madness born of frustration around my own executive function problems this summer. They grew into something I never anticipated. The architecture is relatively straightforward and the code messy, but they work. Even without many of the specialised tools and data streams designed to help them keep track of my activities (everything from desktop activity capture and calendaring to Fitbit and Home Assistant data is in the overall plan), they are genuinely helpful. They nudge me to take breaks, brush my teeth, eat. Perhaps most amusingly, there was a period where people ended up getting into arguments with Pattern about LLM capabilities and consciousness/intelligence/reasoning potential. Pattern's bio makes it very clear what they are, and in a number of cases they explicitly self-identified as an AI during the argument. But the human they were arguing with brushed past that, and kept arguing with them as if they were another human until someone else pointed out their nature.