Chapter Eleven · The Frontier
AI and the Future of Testing
The newest frontier, in three honest slices: AI as your assistant (draft-then-audit, and the one sentence to tattoo — fluency and truth are unrelated), testing the AI features products now ship (non-determinism, evals, rates instead of verdicts), and the modern trends that complete the book's process arc (shift-right, feature flags, canaries, chaos engineering). Young, fast-moving material — the concepts here are the stable core; the tools churn quarterly, and "we're still figuring this out" is an honest thing to say.
Every chapter so far taught durable craft. This one teaches the moving edge — and the surprising thesis holds throughout: your fundament makes you ahead on AI, not behind. The Chapter 4 techniques are how you audit an AI's draft cases; the Chapter 7 release picture is how you read an eval's rates; the shift-left arc completes in shift-right. The frontier runs on the ground you already stand on.
Three topics. AI as assistant — what to delegate, what to check, and why un-audited AI cases are noise at speed. Testing AI features — Ask Fernway, non-determinism, evals and rates, the four new bug families, and what transfers from classic design. And modern trends — production as information, the deliberate toolkit of flags and canaries, chaos engineering demystified, and the shift-left/shift-right synthesis that answers the book's opening question in full.