Long-form research for decisions that need evidence, traceability and an accountable judgement.
REVIEW / SPEC KIT / FREE
GitHub's Spec Kit says its constitution acts as a compile-time check. We fed it 37 known-bad changes to find out what it refuses.
REVIEW / OPENSPEC / FREE
OpenSpec runs a change lifecycle ending in a validated archive. We fed it 37 known-bad changes, then asked if the work was done.
REVIEW / BMAD-METHOD / FREE
BMAD-METHOD says its reviewer never pads. We fed it 37 known-bad changes, then aimed its adversarial lens at working code.
Interviews, registers and essays. Free.
FULL REPORT / THREE HARNESSES / PAID
The same 37 known-bad changes past Spec Kit, OpenSpec and BMAD-METHOD, plus one build through each, to see what each one refuses.
63 scenarios across four configurations, what survives everywhere and what closes it, the three patterns only visible side by side, and the board read on what the harness bought.
Make your agents work.
