red/green TDD

Overview

red/green TDD is a test-driven development pattern adapted for coding agents. In this workflow, an agent writes tests first, runs them to confirm they fail (the “red” phase), then implements code and reruns the tests until they pass (the “green” phase). In the AI coding context, this creates a disciplined feedback loop that helps agents stay grounded in explicit requirements instead of generating code that only appears correct.

For AI Product Managers, red/green TDD matters because it turns agentic coding into a more measurable and reliable process. Rather than asking an agent to “build the feature” in one step, teams can structure work around verifiable outcomes, reducing silent failures and making iteration easier to supervise. It has emerged as part of a broader set of agentic engineering patterns associated with modern coding agents and lightweight project scaffolding such as “thin templates.”

Key Developments

• 2026-02-23: Simon Willison’s Agentic Engineering Patterns was highlighted as a hub for building and operating agentic systems, with red/green TDD called out as one of the coding-agent patterns it collects.
• 2026-04-03: A newsletter mention explained the operational prompt behind red/green TDD: instruct agents to write tests first, run them to confirm failure, implement the code, and rerun tests to confirm success. This framed the pattern as a practical workflow for tools like Claude Code and GPT-5.4.
• 2026-04-04: Lenny Rachitsky shared Simon Willison’s view that late 2025 marked an inflection point for AI coding, with autonomous coding agents, benchmark progress, and red/green TDD combined with “thin templates” helping make agentic software development more dependable.

Relevance to AI PMs

• Specify work in acceptance-test form: PMs can translate feature requirements into concrete tests or failure conditions, giving coding agents a clearer target and reducing ambiguity in execution.
• Improve reliability in agent workflows: Red/green TDD creates a built-in validation loop, which is useful when shipping AI-assisted code changes that need stronger guardrails than one-shot generation.
• Measure agent performance more concretely: Teams can evaluate coding agents based on whether they produce failing tests first, close the failure gap, and reach a passing state consistently across tasks.

Related

• Simon Willison: A key source associated with documenting agentic engineering patterns, including red/green TDD for coding agents.
• Lenny Rachitsky: Amplified the idea by sharing Willison’s framing of the AI coding inflection point and the growing importance of this workflow.
• Claude Code: One of the coding-agent tools mentioned in connection with using red/green TDD in practice.
• GPT-5.4: Referenced as part of the new generation of coding agents capable of following red/green TDD workflows more reliably.
• agentic-engineering-patterns: The broader pattern library where red/green TDD appears as a specific operational technique.
• test-driven development / red green test-driven development / red-green TDD: Common aliases and adjacent formulations of the same concept, adapted here specifically for AI coding agents.