Privacy adoption at scale must work backwards from preserving utility. This research is a series of progressive experiments exploring decentralized software technologies to uncover primitives that preserve utility while strengthening user ownership. Successful ideas from these explorations will eventually be adopted into Tiles.
Active Projects
May 2026WIP
Local models meet AT Protocol
Experiments with ATProto Lexicon schemas, PDS, schema translation, and generative UI for local AI assistants.
Read research →
May 2026WIP
Inference Time Memory Module
Simple inference-time memory module that treats memory management as a series of LLM calls and agent loops over a markdown-based file tree.
Read research →
Archived Projects
Mar 2026EXPERIMENTAL PROTOTYPE
Memory Model Trained with Online RL
On-device memory model trained with online reinforcement learning.
Read research →