Live / Cloudflare D1 release intelligence
Shipstone
A small intelligence pipeline that watches AI and mobile engineering sources, stores every important decision in D1, and publishes practical briefs to the web, Bluesky, and Telegram.
Evidence
Proof before polish.
- Public readers render from D1 with KV reads disabled.
- Source runs, failed fetches, AI calls, quality events, budgets, social delivery, and retention runs are persisted.
- Weak or stale summaries are rejected before public/social delivery.
What it proves
The useful part is not the summary text alone. The system proves that every source attempt, relevance decision, quality rejection, budget check, and social delivery can be inspected after the fact.
Why it matters
Small AI products fail when the output looks good but the process is invisible. Shipstone keeps the process visible, so bad posts can be removed, stale prompt versions can be blocked, and source coverage can be improved without guessing.
Next improvement
The next product work is public presentation, not more pipeline sprawl: sharper showcase pages, useful articles, and a launch pass for metadata, links, accessibility, and performance.
Stack