☕️🧘 Same workflow.
22.5× → 101.
1 min read
Originally posted on LinkedIn · 30 reactions · 1 comments · View original →
☕️🧘 Same workflow. New ceiling. 22.5× → 101.5× in 4 months.
- In Aug/Sep last year: I measured ~22.5× vs. my pre-AI baseline on net code merged through strict quality gates.
- Last 2 weeks: 101.5× → small pod, real product, strict quality gates, production-grade code. ✅
The weird part: my workflow didn’t change. 🤯 Same build-measure-learn cycles with human-in-the-loop. Same agent harness.
What changed was the stack underneath it:
- Opus 4.5 quietly levelled up (Dec-Jan: longer horizon, stronger task execution)
- Claude Code improved as the orchestration layer (planning + execution ergonomics, control, flow)
This showed up while I was refactoring Euda, applying learnings from OpenClaw’s architecture (memory, first-class tool calling, skills, reactive/proactive loops, heartbeat/cron, etc.).
Next measurement round: Opus 4.6 + Agent Teams. Curious to see how much the ceiling moves again. 🚀