The site had no way to show what a year of building actually produced. The Chronicle covers that as a story — eight chapters, narrative pacing, the works. What it does not do is answer a plainer question: what shipped here, and what did it teach. This window is the site going live, and it lands three things at once.
The Chronicle is the long form. Eight chapters from the first embers through three launches, each backed by its own source docs.
The references library is the evidence layer. Every citation in a chapter resolves to a page here. The redaction rules run first, so what publishes is already free of personal details and internal hostnames — nothing is hand-checked at the end and nothing slips through.
The contact scoping wizard is the practical one. Kind, size, integrations, urgency → an estimate range and a recommended engagement, attached to the brief the visitor writes. The optional AI polish path is stubbed; the wizard is complete without it.
Underneath all three, the wireframe persona is gone. Invented biography, invented clients, an invented exit — deleted outright, along with the script that generated them. Site identity now comes from one config file, so there is exactly one place for it to be wrong.
Lessons learned
- Fabricated content is worse than no content. Editing fake history down leaves residue and invites the question of which parts survived. Deleting it and writing the real thing is faster and cannot embarrass anyone later.
- Redaction belongs in the pipeline, not in the prose. If the rules run as a step over the source docs before publish, nobody has to remember to scrub a particular paragraph. The failure mode of a manual check is that it gets skipped.
- An estimate the visitor produces themselves beats a contact form. It gives the conversation a starting number and filters out the briefs that were never going to fit. The form still exists — it just arrives with context attached.