Thoughtworks warns developers hesitate to fully trust coding agents in production
Thoughtworks’ senior technologists explain why experienced developers remain cautious about agentic development. Their hesitation stems from deep concerns about reliability, guardrails, and the gap between building code and ensuring it works safely in production. The firm’s internal experiments showed agents often produce code faster but fail to meet stability standards, with continuous…
Key points
- Experienced developers distrust agentic coding due to unproven production reliability and lack of guardrails
- Thoughtworks’ experiment found agents produced code faster but caused frequent red builds and instability
- Teams spent ⅔ of time designing feedback systems, not writing code, highlighting gaps in automation
The team’s week-long experiment—where agents handled all coding—revealed developers spent two-thirds of their time designing boundaries, feedback loops, and testing frameworks rather than writing code. While agents generated drafts, humans had to manually refine architecture, contracts, and delivery pipelines. Thoughtworks’ Giles Alexander noted that even after two iterations, teams hadn’t resolved core challenges like batch size, test brittleness, or parallel agent conflicts. External data from DORA’s 2025 research aligns with these findings: AI adoption correlates with higher delivery speed but lower stability, reversing decades of DevOps progress.
Why are good developers so wary of agentic development?
martinfowler.com · 29 October 2026
Loading the full article…
This text was published by martinfowler.com. It is reproduced here with attribution so you can read it in full; the rights remain with the publisher. Read it at the source ↗
Coverage and discussion
1source- Hacker News discussion · 43 pointsnews.ycombinator.com
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Agents & Tools
All →- Claude Code finds all 5 planted vulnerabilities in 8 minutes · 1 src
- MoEngage uses Amazon Bedrock AgentCore Gateway to expose 400 internal APIs as agent tools · 1 src
- Wabi raises $55M at $1.25B valuation to expand personal AI assistant · 3 src
- Ben Hylak says every product should be a headless mcp feeding an agent · 1 src
- Amazon adds 100 shopping features to Alexa for lists and subscriptions · 2 src
Comments
via GitHub Discussions