Exploring what remains irreducible in the age of Generative AI

I build for the agency of individuals, and for the worlds we share with AI agents and other creatures. Underneath both, I build the systems that make them possible.

Ownership lives in the architecture.

Personal AI should keep information flow, memory, orchestration, and delivery inspectable and portable, so models, providers, and tools can change without taking control away from the person.

Amanuensis

attention · aggregation · signal extraction

A policy-gated information triage layer for personal LLM assistants. It gathers messages, tasks, events, and ambient feeds, then uses context to rank, denoise, and route each signal, so attention goes to what matters, when and where it matters.

Yaaa

Ops architecture · conceptual framework

A self-hostable personal-assistant architecture that keeps memory authority and side-effect control with the user while models, harnesses, and tools remain replaceable.

Enablement, not replacement.

Technology should lower the barriers of memory, attention, and mobility, and leave every decision in the person's hands.

Beagle

WIP

A cognitive-support agent concept for people facing memory or attention barriers: an always-on assistant that logs, tracks, and reminds, so the person can make informed decisions even when being informed is a challenge.

Trailwalk

working MVP · field media library

A working VR prototype built from real 360-degree trail footage and spatial audio, bringing outdoor places into homes and community settings when the trail is hard to reach.

Worlds are made of relationships.

Relationships among people, AI agents, machines, and other beings shape how a world is shared and experienced, irreducible to any outcome.

commonplace

Research WIP

Research into human-AI relationships in game worlds, curiosity-driven autotelic agency in AI agents, and the philosophy of Otherness as it emerges from how we treat and regard LLM agents.

If this overlaps with something you're trying to build, study, or care for, I'd be glad to talk.