Every prompt behind Woodstar, session by session. The app is AI-written; these are the instructions it was written from.

The policy — what's published, what's redacted, what's not here

Woodstar is largely built with AI, and this is the provenance record. Each session below lists every prompt typed into it, in order, with timestamps. Secrets are the only thing removed: credentials and identifiers that invite abuse are redacted in place, and every redaction is annotated with why (legend below). ⟦redacted: …⟧ marks are counted in the open — the published count is what keeps redaction honest.

What's not here, said plainly: the model's replies and the commands it ran. Eleven Messenger publishes its sessions whole; Woodstar publishes prompts-only until it has the same redaction machinery, because a transcript can leak what a prompt cannot. The prompt texts themselves are verbatim — typos and all.

[Image #n] marks a screenshot pasted into the prompt; images aren't published.

What gets redacted, and why

Prefer the story to the log? Milestones — the key moments of building Woodstar, each linked to the session where it happened.

Sessions

Local transcripts prune after roughly a week, so the ledger accrues from each run of scripts/status-tally.mjs forward. Newest first; for the story in order, see Milestones. Search matches summaries and every prompt inside each session.

WhenModelSummary