scottparker

I design how people learn. And I build the systems that do it.

Program strategist and learning designer who has shipped enterprise learning programs and platforms, and now builds AI systems for learning and content, with a documented method for doing it well.

See the AI workContact

SPECIFICATIONEXECUTIONPRODUCT OWNERa human: sole decision authorityapproves every batchrules on every forkARCHITECT · Claude in chatoptions, specifications,briefings; never decides,never buildsBUILDER · Claude Codeexecutes against the specification,in PR-per-branch increments;escalates ambiguity;never touches governanceThe role is tool-agnostic: another codingagent, or a project’s own pipeline, couldsit here. Claude Code does today.rules onforksbriefing withacceptance criteriaREVIEW FROM SOURCEdiffs, raw test output, CI logsnever builder self-reportPR per branchescalates ambiguitydisposition to the ownerGOVERNANCE lives in the repodecision log · manifest · deferred-items registerthe owner applies every record; the builder never touches it
From the AI-Assisted Development Methodology: the specification → execution → review seam. Read the case →

How I work

Specify before you build. Options, not conclusions. Evidence over assertion. Files, not transcripts. Humans at the center, not just in the loop. More →

Earlier

Educational science games for grades 4–8 · an early avatar-based online community · Flash-era rich media · a Brandon Hall Silver for Cisco. Archive →