About wwwatch

What it is

wwwatch is a daily AI journal for product engineers. Every day, a pipeline collects what moved in AI — new model releases, framework updates, research papers, funding rounds, security incidents — scores it by editorial signal, and writes 300-500 word articles from the actual source content.

No hype, no summaries of summaries. Each article cites its source and only reports what the source actually says. A human reviews every draft before it goes live.

The pipeline

1.Collect — Hacker News, GitHub releases, Hugging Face papers, Reddit, RSS
2.Score — event_freshness × 0.45 + authority × 0.25 + engagement × 0.20 + keywords × 0.10
3.Enrich — fetch actual source content (release notes, papers, articles)
4.Write — one Sonnet call per article, constrained to fetched source material
5.QA — flag short articles, missing sources, unsourced factual details
6.Store — draft edition in DB, human review before publishing

Built by builders

wwwatch runs on Next.js, Neon Postgres, and Claude Sonnet. The pipeline is a deterministic cron job — no live LLM calls in the request path. Pages are server-rendered from the DB. Costs are bounded and independent of traffic.