Experiment Suggests TDD in AI Coding Agents Might Be Just Theater
hichaelmart · x · 2026-08-12
Birgitta Böckeler, Distinguished Engineer at Thoughtworks, explored the actual value of integrating Test-Driven Development (TDD) with AI-assisted coding.
The article outlines three ways to combine TDD with AI: humans writing tests for AI to implement, AI writing tests for human review, and fully autonomous TDD inside the agentic loop. While the fully autonomous approach is currently the most popular, the author ran an exploratory evaluation to test its effectiveness.
The preliminary results showed no clearly discernible difference in quality when using a TDD workflow versus not using one. In fact, in several instances, the non-TDD workflow solutions were ranked slightly higher in design by the evaluating model (Opus), suggesting that practices good for humans might be irrelevant or bad for coding agents.
More from coding & agent
- UX Pain Point in Coding Agents: Users Call for Progress Time Estimates — Turbulent_Ad_1039 · 2026-08-12
- Building an AI-Native 'Software Factory' with Agents for Automated Deployment — Pavan_Belagatti · 2026-08-12
- fastdup: Open-Source Tool for Image & Video Dataset Deduplication — tom_doerr · 2026-08-12
- Building an AI Accounting Agent for 2026: What If There's No Existing Workflow? — RajaswaPatil · 2026-08-12
- DHH merges PR to bypass permissions for coding agents in omarchy, says treating AGI like a toddler is wrong — mitsuhiko · 2026-08-12
- Two Agents on One Repo: My Setup Failed, What's Yours? — Specialist_Agent3599 · 2026-08-12