Start here for what this section covers and how changes reach production—the articles below drill into confidentiality, site automation, early operations, and open design questions. What you read is the public website—the window into how the automation works—not a clone of TB’s private repositories or the MagnetMagpie build repository.
The delivery loop is intentionally mechanical. A PM agent works from Linear tickets and skills in this repository; builder agents implement changes, run checks, and open pull requests. CI runs bash build.sh (content pipeline, Hugo, Playwright). When checks pass, auto-merge lands qualifying cursor/* pull requests on main; Deploy Labs Site publishes when AWS is configured. Humans set process and review at gates—they are not in the routine merge-and-deploy loop for this repo. Schedules slip, workflows fail, and we rewrite skills when something breaks: treat this as an evolving experiment, not a finished product. For how TB builds financial tools in Trunk and related repos, see How we engineer.
Where to read next: How this site handles confidentiality explains the trust boundary and what may appear on the Labs site. For tools, workflows, and the path from ticket to production, see How this site builds itself. This week on main is a rolling twelve-week log of merged pull requests from git history at build time. For a timestamped account of day-one failures and how the loop responded, read Day 1 live replay. The human side of early operations describes what the human role looked like during those first weeks. The open questions name design tensions we have not settled. Annotated artefacts walks key files in the repo.
Live status
Live operations is a build-time snapshot of recent activity, coverage, and open questions—what the pipeline knew when the site last built.
Articles in this section
The list below picks up where this overview leaves off—you can read in any order.