Repository: github.com/avery-shaw-demo/ai-augmented-portfolio
AI-collaboration evidence
Heavy use of Copilot and Claude Code visible in PR descriptions. .cursorrules in 4 of 6 active repos. Velocity is real: 87 PRs merged in 90 days across personal repos. Pattern fluency with prompt-engineering.
Code craft
Quality varies by repo. The flagship repo has tight types, named errors, and isolation between modules. Two side repos have copy-paste between files (same helpers duplicated 3x), inconsistent error handling, and 'TODO: fix this' left in shipped code. Inconsistency is the craft signal worth verifying.
Design ownership
Hard to tell from artifacts alone whether design decisions are owned by the candidate or generated. Architecture diagrams in /docs read polished but haven't been updated to match the last 3 weeks of code. Interview question: walk me through why you chose X over Y.
Synthesis
Output volume is high; AI-tooling fluency is real. Code-craft signals are uneven across repos — interview should probe ownership of the harder design decisions.