Kent Beck: Trust, Not Code Generation, Will Define Engineering in the AI Era

Kent Beck — creator of Extreme Programming, co-author of the Agile Manifesto, inventor of TDD — has been writing code for over four decades. His latest argument: the skills that matter most in AI-era software engineering aren’t about generating code faster. They’re about building trust.

Beck’s career arc maps directly onto the shifts AI-coding practitioners are navigating right now:

• TDD isn’t dead — it’s more relevant. Beck sees test-driven development as a trust mechanism, not a productivity trick. When AI generates code, tests become the contract that tells you whether the output is safe to ship.
• Agile was about feedback loops, not ceremonies. The original insight — short cycles, working software, direct communication — applies directly to how developers should interact with AI agents. Prompt, evaluate, adjust. The loop is the method.
• Design still requires human judgment. Beck recently caught flak on LinkedIn for keeping all his Rust B+ tree code in a single file. His response: “I know what I’m doing. I just don’t know when I’m done.” Modularity is context-dependent. So is every design decision an AI agent makes on your behalf.
• Software engineering is a social activity. Beck’s current book project focuses on how teams practice design together while shipping features — the part AI can’t automate.

Beck has been exploring what AI changes about software development — and what stays the same. He spoke at O’Reilly’s Coding with AI seminar on vibe coding, and appeared on The Pragmatic Engineer podcast discussing TDD and AI agents. His consistent position: AI amplifies existing engineering discipline. Without it, you just generate bugs faster.

The takeaway is specific. If you’re using Cursor, Claude Code, or any AI agent to write code, Beck’s decades of work point to one practice worth adopting now: write the test first, let the AI write the implementation, and treat the test as your source of truth — not the generated code.