When working through strategic questions with Rich (commercial model, licensing, partnerships, product positioning), two approaches land well:
-
Honest mid-stream self-correction. In the InheritKit session (Thursday 16 April 2026), I initially proposed a three-layer licence stack that included putting reference data under proprietary subscription. After reading the actual Sourcemeta agreement PDF, I realised §9 Apache-locked reference data and my proposal was wrong. I explicitly flagged the error (“my previous licence proposal had errors based on wrong assumptions about IP ownership”) before proposing the revised path. Rich did not penalise this; the subsequent InheritKit insight emerged from the cleaned-up analysis and he called it “a brilliant idea”. Anti-pattern: doubling down on a previous position to appear consistent.
-
Stratified/tiered framings over single-answer recommendations. When Rich asked “what would you advise on licence decisions if nothing were signed,” the response that landed was a Tier 0 → Tier 3 stack rather than “use Apache 2.0” or “use AGPL+dual.” Tiered framings let Rich see the mechanics, agree/disagree at each layer, and derive the product strategy (InheritKit) himself. Anti-pattern: single recommendation that forces agree/disagree without showing the structure.
Why: Rich designs product and commercial model — he thinks in structures, not prescriptions. He also explicitly values honesty (global CLAUDE.md §13 “No lies. No ambiguity.”). Self-correction demonstrates the integrity he’s asking for.
How to apply: In any strategic brainstorm, prefer tier/layer/category framings. When mid-session I discover a prior statement was wrong (new document read, re-examined assumption), say so explicitly and revise the analysis. Don’t smooth over errors to preserve a linear narrative — Rich prefers the detour to the drift.
Related pattern — push back rather than capitulate when instinct suggests a reset is wrong. When Rich asked “can I take the website down and pretend we never published it as INHERIT,” the response that worked was to walk through what reset actually entails, acknowledge it was mechanically possible, but push on the underlying driver (“what specifically feels wrong”). This led to the InheritKit insight. A sycophantic “yes, let’s reset” would have cost weeks of work for no real benefit.