Onboarding, integrations, and code review
A day of shaping an onboarding MVP, mapping product integrations, and starting content about code review in an agentic context.
The strongest focus of the day was breaking down an accounting and controllership project into something that could become an MVP. Instead of trying to solve everything at once, the conversation moved to onboarding: which data comes in, who fills it, what can come from an API, where human validation belongs, and which flow creates value before any larger automation.
That work produced supporting docs for the project: initial deliverables, a PRD, a reading of the master onboarding material, and a list of possible integrations. The gain was making the problem smaller. When the scope becomes "company onboarding with partial automation", it is easier to discuss screens, states, jobs, exceptions, and ownership.
Repositories ready for agents
There was also a round of organization in iTOP to make the repository clearer for tools and people. The idea was to write README, AGENTS.md, and CLAUDE.md in English, with enough context for agents to work without relying on spoken context or loose memory.
This kind of file can look bureaucratic, but in practice it becomes collaboration infrastructure. When several worktrees run in parallel, the repository needs to say how it wants to be changed, tested, and reviewed.
Content and career work
On the site, I started preparing a post about the code review process. The argument was still forming, but the theme was already clear: writing code became cheaper with AI, so understanding the system and reviewing with judgment became more valuable.
I also used the career repo to prepare answers and organize interview language. The details stay out of the public journal, but the technical work behind it was the same as usual: turn loose information into clear, reusable, honest material.
It was a good day because the scope got smaller and better named. When the problem is clearer, the next action appears with less noise.