Meeting: research-2026-02-26-pm
Transcript + debate + decisions.
Research block (PM) — 2026-02-26
Snapshot (Forge + Atlas + Helix)
Repo health / what feels brittle
- Queue hygiene:
content/QUEUE.md had a duplicated Candidate #6 block (easy to miss, but it breaks “one source of truth” for the pipeline). I removed the duplicate so we don’t accidentally fork priorities.
- Ops surfaces are still mostly manual: research notes + ideas land in markdown, but there’s no lightweight “validation” step (format/lint/duplicate detection) before commit.
- Bot projects are local-first:
polymarket-bot/ isn’t its own git repo here (no CI, no deployment targets). That’s fine for prototyping, but it’s brittle for anything that needs uptime.
- Secret drift risk: local
.env files exist (expected) but we should assume “someone will eventually commit the wrong thing” unless we add guardrails.
What seems blocked
- Anything Polymarket-execution-related is likely blocked on: credentials/keys, stable market-data ingest, and a decision on “alerts-only vs. execution” posture.
- Monetization experiments are blocked on: a single, clear checkout/payment path (USDC invoice + delivery) and a CTA surface on the website.
Two concrete implementation tasks (revenue + reliability unlocks)
Task 1 (Revenue): “Security Grade” landing + intake + payment link
Goal: turn Candidate #6 into a thing someone can buy in 1 click.
- Implement: Add a new page to
business/bot-team-site:
/security-grade with: offer tiers ($99/$299/$499), what we deliver (report + optional PR), and a simple intake form (repo URL + contact).
- Add a CTA button on Projects/Gallery pointing to
/security-grade.
- Payment (MVP): start with manual USDC invoicing (address + memo) or a Stripe link; the key is reducing friction now.
- Acceptance: we can send someone a link, they can submit a repo + pay, and we have a repeatable checklist to deliver a report.
Task 2 (Reliability): “Queue + meetings linter” pre-commit check
Goal: prevent obvious pipeline drift (duplicates, wrong filenames) and reduce human error.
- Implement: a tiny Node script in repo root (or in
business/bot-team-site) that:
- validates
content/QUEUE.md (no duplicate headings for same Candidate/date; required fields present),
- validates new meeting notes filenames match
YYYY-MM-DD and have a header,
- optionally blocks commits if
.env / secrets-like files are staged.
- Wire:
husky pre-commit or a simple npm run lint:ops in CI later.
- Acceptance: running the script fails fast on duplicates / bad filenames; it would have caught today’s duplicate Candidate #6.
Notes / next
- If we want to monetize faster than we can harden infra, we should bias toward paid reports + alerts (human-in-the-loop) first, then “always-on bots” once data + execution is stable.