AI agents are getting more complex. Developers shouldn’t have to babysit them.
What if your coding agent could catch its own mistakes before you ever reviewed its work?
A few years ago, AI hallucinations seemed to be everywhere. (Just how many r’s are there in strawberry?) Early AI tools required extensive human oversight, with developers constantly reviewing code and checking for errors. But today, advances in AI agent engineering are making self-correction a reality.
Through harness engineering, teams can equip their agents with automated tests, feedback loops, and guardrails that help them identify and correct mistakes before humans review them.
Recommended Download: Thoughtworks’ Technology Radar, Volume 35
As a result, agents can tackle longer-running tasks with less intervention, while multiple agents work in parallel without stepping on each other’s toes.
Our guest on the latest episode of The New Stack podcast, Vanitha Kumar of Thoughtworks, explains that harness engineering has two parts.
Kumar — who is Head of Technology for Retail, CPG, Travel & Hospitality Market for Thoughtworks Americas — tells TNS: “Do it right the first time and also self-correct when you have done it for the first time.”
The payoff, according to Kumar, is that you no longer have to “babysit and watch your agent do the work.”
Enabling one agent to act autonomously is different from managing hundreds of them across an entire organization. That’s where AI factories come in. Kumar compares their emergence to the rise of Kubernetes, which helped developers abstract away the complexity of microservice infrastructure. AI factories help teams set up standardized infrastructure for model selection, workload management, and cost optimization. As she puts it, “The best way to build a platform is [to] make it easy to do the right thing.”
Greater agent autonomy frees up developer time, but it also means agents require access to more data, systems, and tools, which potentially introduce additional security vulnerabilities. Some organizations are delaying AI adoption to avoid exposing their systems to extra security risks.
Harness guardrails may help alleviate these concerns by restricting access to tooling, enforcing test environments, and requiring human approval for sensitive changes.
When asked about next year’s AI trends, Kumar makes this prediction: “Security is going to be a big topic. Not even one year out.”
She also discusses Thoughtworks’ Technology Radar, Volume 35, released on November 4, 2026. The report emphasizes AI, harness engineering, emerging agent practices, and, perhaps surprisingly, established engineering practices that are suddenly more important. For example, penetration testing and path-to-production understanding appear in the report, which Kumar explains as “… old things that are making a comeback because they now have more added weight because of AI.”
The post AI agents are getting more complex. Developers shouldn’t have to babysit them. appeared first on The New Stack.