full time… i work in product
in my spare time… i build small systems to understand larger shifts in software
software used to be about workflows
it is slowly becoming about judgment
from buttons to context
from forms to memory
from interfaces to decisions
that shift is what i like exploring
- what software looks like when it remembers
- how agents coordinate… not just respond
- where intelligence actually reduces friction
- what makes ai useful inside real workflows
- tetris.codes - a small program between your coding agent and the model. context shrinks on your laptop — six to eight times fewer tokens leave it (tetris.codes)
- nxl — programming language that reduces LLM token consumption by 70% while maintaining full semantic equivalence
- nexus prime — local first control plane for coding agents (nexus-prime.cfd)
- mindkit — capability layer for multi agent systems
- grain — interaction layer for ai interfaces
- cipher.sys — self destructing to do app
- knight — privacy first chrome extension for local resume parsing
- phantom knowledge layer — pkl - vector embeddings, rag, & connectors for phantom pm os
the next generation of software will not just execute
it will participate