lyndonkl Update Watchlist
Workflow
Monthly (first weekly run of the month):
- [ ] Step 1: Read last 4-8 weekly digests' appendices
- [ ] Step 2: Identify sources with ≥3 consecutive fetch failures (remove candidates)
- [ ] Step 3: Identify external authors/URLs referenced by watchlist sources ≥3 times in a month (add candidates)
- [ ] Step 4: Parse any user-added markers in digest bodies (<!-- scout: drop-this-source --> / <!-- scout: add {url} -->)
- [ ] Step 5: WebFetch each proposed add to validate it returns content
- [ ] Step 6: Emit ops/trend-scout/watchlist-proposed-diff.md
- [ ] Step 7: On user approval (.approved file or explicit trigger), apply diff; append changelog to watchlist.md
Proposed diff format
# Watchlist changes proposed 2026-W17
+ ## people
+ - name: "Interconnects (Nathan Lambert)"
+ url: https://www.interconnects.ai/archive
+ source_type: blog
+ essential: false
+ rationale: "Referenced 6x in last 4 weeks from Raschka, Willison, Transformer Circuits."
- ## blogs
- - name: "someblog"
- url: https://someblog.example/
- source_type: blog
- rationale: "4 consecutive fetch failures."
Guardrails
- Never add without WebFetch validation.
- Never remove
essential: truesources without explicit user approval. - Cap adds at 3 per month. Prevents drift.
- Preserve clustered structure (blogs / people / research / aggregators / domain / social).
- Changelog entry appended to watchlist.md on every apply.
- Monthly only. Don't run weekly.