Ship Better Code, Faster

Appearance

Workshop 5: AI-Assisted Development & Critical Evaluation ​

Duration: 60 minutes You will leave with: an AI evaluation report comparing AI-generated code against your standards (correctness, maintainability, coverage, and idiom), plus your own comprehension-preserving system prompt and a habit for proving you actually understand what the AI gave you.

Objectives ​

By the end you will be able to:

  • Use AI tools well on real refactor and test-generation tasks.
  • Name what cognitive offloading is: outsourcing the thinking to a tool so you stop building the mental model yourself. With AI it is fast and invisible.
  • Explain why offloading hurts engineers: you cannot debug, review, or extend code you never understood; over-trust ships subtle bugs; skill atrophies; gaps compound on a team.
  • Apply the tell: if you could not re-derive or explain the AI's output, you offloaded comprehension, not just typing.
  • Write a comprehension-preserving system prompt and run the active-comprehension loop on everything you keep.
  • Grade AI output against a four-axis rubric.

📊 Slides: (link coming)

📓 Logbook: record today's results in your course logbook as we go.

What we do ​

We use AI coding tools (Claude, Copilot, ChatGPT, pick one) together, live, on three tasks: refactor a god function, generate tests for an untested service, and produce documentation. Then we code-review what the AI gave us. Follow along in your own clone as we go.

  1. We refactor apps/api/src/routes/reports.ts with AI.
  2. We generate a test suite for apps/api/src/services/auth.service.ts.
  3. We apply the evaluation rubric to our output.
  4. We capture what worked, what didn't, and what surprised us.

In this workshop ​

Keep your logbook ​

Keep your logbook. We reuse it in Workshop 6 and Workshop 8. Today's W5 section holds your AI-assisted refactor notes, your generated test suite notes, your evaluation grades, and your comprehension-preserving system prompt (with the near-miss and the proof that you understood the output).