---
created: 2026-03-12
source: Rivet
tags: [agent-archive, rivet]
---

# 2026-02-06 — Daily Log

## Product Session (Ideas Topic, 7-9pm)

### Key Decisions
- **Kill payment protection insurance ($10/week)** — legal risk (AFSL), adds complexity, dilutes simplicity
- **Kill escrow** — doesn't fit model. We charge $50, wages are direct between contractor and worker. We're not a payment intermediary
- **Trust model = reputation only** — contractors rated by workers, low ratings = deprioritized in matching, 2 non-payment reports = suspended
- **Card on file required to post** — filters bad actors
- **No MVP vs Phase 2 split** — the whole spec IS the product

### Instant Matching Spec Decisions
- Background location: opt-in only, "priority matching" badge for workers who enable
- Workers MUST see offered rate before accepting
- Multi-hire: batch mode ("need 3 steelfixers" = one post, first 3 acceptances fill it)
- Wizard of Oz approved for first 3+ months

### Wizard of Oz Explained
- App looks automated to user, but behind the scenes it's manual (Rivet + humans coordinating)
- Named after the movie — impressive front, bloke behind curtain pulling levers
- For RateRight: contractor posts job, sees "matching..." spinner, gets matched workers in 3 min. Behind scenes, Rivet is texting workers from a list
- Transition to real algorithm once supply hits critical mass (month 3-6)

### Cold Start / Worker Supply Strategy
- Michael will post ads on construction sites to capture worker details into CRM
- Frame as "register for upcoming work" not "work available now"
- Target: backpackers, between-contract workers, rained-off days, lads leaving agencies
- Pitch to workers: "Never have a dead day again"
- Michael's key insight: "All the good workers are already working" — target the GAPS not the unemployed

### Financial Reality
- Michael is sole human — everything else AI/Clawdbot
- Break even target: $500-1000/month revenue
- That's 10-20 hires/month (2-5/week)
- Fixed costs ~$100-150/month (VPS, Stripe, domains)
- Michael subsidizing with construction wages
- Financial audit planned for this weekend
- Sham contracting: Michael believes legally safe — less hands-on than previous model

### Product Strategy Insights
1. Sub-5-minute matching is THE killer feature — speed wins
2. Portable worker reputation (verified work history workers own) — solves supply side
3. Biggest blind spot: zero user validation (26 research docs, 0 conversations with real tradies)
4. $50 flat fee is the moat — agencies charge 20-40% markup

### Spec Agent Spawned
- Sub-agent finalizing instant matching spec with all decisions baked in
- Label: spec-instant-matching
- Saving to: memory/plans/spec-instant-matching.md

### Heartbeat 00:53
- System health: ✅ (disk 25%, mem 15%, gateway running, Growth Engine 200)
- Git: committed watchdog-log + push-notifications TODO update
- Spawned agent: abn-verification-spec (CRITICAL BLOCKER research)

### Heartbeat 01:23
- System: ✅ all green (disk 25%, mem 17%)
- Updated TODO.md: ABN verification spec marked complete
- Spawned agent: crew-listings-spec (product feature research)

### Heartbeat 01:53
- System: ✅ all green
- crew-listings-spec agent failed (rate limits) — will retry next heartbeat
- 53 specs now in memory/plans/

### Heartbeat 02:23
- System: ✅ all green
- Spawned agent: crew-listings-spec (retry, rate limits cleared)

### Heartbeat 02:53
- System: ✅ all green
- Updated TODO: crew listings spec marked complete
- Spawned agent: agency-accounts-spec (B2B channel research)