ethos
Running Record

Log

Decisions, learnings, and progress across the incubator. Updated weekly.

📋

Methodology & Frameworks

Evaluation criteria, kill/continue framework, distribution channels

→

Entries

Guildry Phase 1 complete

2026-02-07
guildry
- Built Blueprint: conversational project scoping with phase suggestions and hour estimates - Built Bench: talent network management (employees, contractors, referrals) with skills tracking - Built Retro: project retrospectives capturing what worked, what didn't, and lessons learned - Full MVP loop now complete: Scope → Staff → Deliver → Learn → Improve - Added optimistic UI for chat, auto-redirect after entity creation

Decisions

  • →Build all three modules before polishing - see full loop working together first
  • →Keep prompts simple for MVP - will iterate based on real usage patterns
  • →Use tool_use content extraction pattern when Claude returns only tool calls without text

Learnings

  • ◆Vertical slices work well - each module followed same pattern: tools → prompt → API → pages → handlers
  • ◆AI tool handlers need to format informational tool outputs (like suggest_phases) into readable content
  • ◆Phase 1 built in ~3 hours across 2 sessions - validates 1 MVP/week pace is achievable

Ethos site launched

2026-02-06
ethos
- Created standalone ethos site to document the Solo AI Product Incubator model - Built /projects page to track active experiments with status badges - Set up /log page (this one) for ongoing documentation - Established weekly log workflow: Monday reminder → Sunday review → publish

Decisions

  • →Keep ethos site separate from guildry and public-brief - each project gets its own repo/deploy
  • →Use markdown-style content in JS for now - low overhead, version controlled
  • →Weekly log cadence with session-start prompts to capture notes throughout the week

Learnings

  • ◆AI-assisted development is 10-20x faster than traditional estimates - Phase 0 took half a day vs 3 week estimate
  • ◆Discovery-first approach beats rigid upfront schemas - let AI conversations reveal what data matters

Guildry Phase 0 complete

2026-02-06
guildry
- Completed foundation: Clerk auth, Supabase with RLS, Claude API integration - AI can create clients via natural language conversation - Fixed schema mismatches (org_id vs organization_id) across codebase - Consolidated documentation into /docs folder

Decisions

  • →Pivot Phase 1 from 17-PR waterfall to discovery-first approach
  • →Let AI conversations reveal what data structure clients actually need before locking schemas
  • →Skip Sentry for now - add before production launch with real users

Learnings

  • ◆Build guides can be overspecified - risk building rigid models that don't match real workflows
  • ◆The 'target schema' pattern (AI extracts structured data via function calling) is flexible enough to evolve