cat ~/ai4se/README.md
## Methodology • Process • Tools • Harmony • Measurement
Notes on how AI4SE / Agentic Engineering reshapes methodology, process, tools, human–agent collaboration, and productivity measurement.
man -k ai4se
Methodology — SDD, Agentic Engineering, AI4SE system
Process — human–agent workflows, review pipelines, loop design
Tools — MCP, Harness Engineering, browser automation
Harmony — HITL/HOTL/HOOL, Planner/Generator/Evaluator
Measurement — DORA, SPACE, DevEx, agent effectiveness
grep -r "featured" ~/notes/
Knowing who owns the contract is not enough. This piece explains local-only, cross-business-repo, and shared-dependency work, with an executable sequence: contract prose in contracts repos, pointer-only impl proposals, parallel work after approval, final pin before merge. An Acme microservice map shows landing, plus a concrete cross-repo walkthrough of openspec /opsx-propose, apply, and archive commands and prompts.
Industry narratives love a single developer-productivity multiplier for AI4SE, but slogans rarely separate belief from measured delivery. This piece names three failures of bare N× claims, proposes a dual-track reading—capability maturity beside labeled throughput and cycle time—and offers a sponsor decision tree plus anti-patterns: estimate ≠ platform-measured, capability lift ≠ ROI.
The most dangerous moment in an AI4SE transformation is often not the absence of results, but the point where local value appears while overall delivery performance temporarily declines. This article uses the DevOps-style J-curve and enterprise consulting experience to explain how Explore, Consolidation, and Scale can turn an experiment into repeatable organizational capability.
Stop framing SDD as a replacement for Agile. Place Agile/Lean, Scrum·Kanban·Scrumban, and SDD on three layers—values, team frameworks, and work-item delivery—so each answers a different question and they compose cleanly.
Scattered prompts are not a User Harness. ai4se-harness turns the SDD playbook’s P0–P5 and G1/G2/G3 into a configurable framework: four work-item Commands as entries, strict/copilot/yolo for autonomy, and three Gate Agents on the stage flow—method and gates stay fixed; asset packs and human–AI depth vary.
In multi-repo or submodule SDD, teams often argue whether specs belong inside one submodule or beside them all. A better test is ownership: shared boundary behavior belongs in a planning layer; per-repo implementation stays local. OpenSpec Stores make that topology concrete.
dpkg -l | grep ^i