Lawyer-built 212k-line payroll SaaS: OpenAI models in the loop saved no Claude quota in testing
Far_Idea9616 · reddit · 2026-09-13
A lawyer with payroll domain expertise shares a deep Claude Code workflow: Fable 5.1 xhigh orchestrates 7-10 step launch sequences while Opus 5 xhigh executes — yielding a 112k-line Python backend, 100k-line React frontend, 183k lines of tests, and 5,300 commits since April. Facing the Max 20x quota cut, he tested OpenAI models: Sol as implementer had integrity failures at high effort and, even at xhigh, external orchestration saved zero Claude budget; flipped to reviewer, Sol xhigh blind-found 2.5/3 known defects plus a live bug; Astra high outperformed Astra xhigh (26% more time, 17% more tokens for nothing). Verdict: it's added complexity — an Astra high reviewer — that may still save quota by catching bugs while context is fresh. Full token-usage tables included.
More from coding & agent
- Anthropic quietly ships server-side rule banning game rewrites in Claude Code — petrusenko_max · 2026-10-31
- Bootstrapping competitive edge with frontier tokens — curious_vii · 2026-10-31
- Element Forge: open-source editable AI VFX for real footage in ComfyUI — zitroffell · 2026-10-30
- Gym-Anything Adds Audit Agents That Try to Break Simulations With Legal Robot Actions — wellecks · 2026-10-30
- durable-object-migrator: zero-downtime exports for Cloudflare Durable Objects — samgoodwin89 · 2026-10-30
- This 5-question interview prompt makes Grok design your first two custom bots in minutes — nateliason · 2026-10-30