Autonomous systems
Not demos. These are loops that fire on a schedule, do real work while I sleep, and leave a full record behind them. The OS schedules them, a pipeline executes, and plain files are the memory.
Reads every stock in the universe each night, scores each informational event against a written rubric, and flips Buy / Hold / Sell ratings by formula. No human in the loop.
Reads ~25 sources, dedupes them into the stories that matter, fact-checks every claim to a primary source, drafts one sharp brief, then grades its own last calls and updates the brain it reads next time.
When armed, each week it reads three days of AI newsletters from the inbox, synthesizes the most interesting threads, finds a non-obvious angle, and drafts a post for manual review. Currently paused.
There is no server and no SaaS dashboard behind any of this. The operating system's own scheduler fires each loop at its appointed time. A small pipeline wakes up, does its work, and exits. The only state that survives is a handful of plain files on disk, which the next run reads before it decides anything.
That is the whole design: the OS schedules, the pipeline executes, files are the memory. It means every run is reproducible, every decision is on the record, and the system stays honest because its full history is right there in the repo. The equity board flips ratings from that record. The newsletter grades its own past calls from it. The LinkedIn loop never repeats an angle because it reads its own history first.
The research board is an educational experiment, not investment advice. Statuses reflect the last time this page was built.