Mapping how an agentic company runs, plus AI ideas for the event platform
Studied how a learning agent infra can serve a whole company, explored AI to improve the event-registration product, and dug into self-hosted deployment trade-offs.
A day of architecture questions more than code. Most of it was thinking out loud with the agent about how the pieces of a small company could actually run on top of an AI infra.
hermes
Spent three sessions working through one core question: how do you run an agentic company where the agent keeps learning, but the company repo only holds skills and area-specific content. The tricky part is that the intelligence lives on the server, not in the repo, so collaborators who clone the company repo locally still need a way to contribute to skills without touching the core. I mapped how the project was built, what it can do today, and how self-created and self-improving skills would fit. Pure design work, no commits.
itop.com.br
Two threads here, merged across workspaces. First, product: I asked the agent to read the whole repo and point out where AI could improve the experience, for both the people who create and organize events and the people who register for them. The angle was product and customer focused first, then technology choices, with onboarding and user docs in scope. Second, positioning: I worked on a clear answer to "what problem does your system actually solve" and saved it to context.
I also went deep on infrastructure trade-offs for this project: whether Nix makes sense here, how the current self-hosted setup compares to managed options, and Docker Swarm versus the self-hosted PaaS approach. Good clarity on what to keep simple.
career
Also spun up a new public portfolio repo and pushed it. Kept it generic on purpose.
Spiritual
Wisdom-literature study on the principle of firstfruits and generosity, saved for later.
Some days the highest-value output is a clearer map, not a green diff.