What Is Loop Engineering? Google Brain Co-Founder Says It Can Make AI Build Better Apps

Artificial

intelligence tools can now write software codes for humans, test them and also offer capabilities to fix them. However, the major question arises how humans can fully trust AI? Andrew Ng, co-founder of Google Brain and Coursera notes that despite AI taking over coding tasks, humans are still important. In his latest open letter, Ng explained how a series of feedback loops allow AI coding agents to write code, test it, fix bugs and continuously improve apps with minimal human intervention.

What Is Loop Engineering?

Ng stated that loop engineering has become a key topic in AI after the term gained attention through social media discussions involving Boris Cherny, the creator of Claude Code and Peter Steinberger, creator of OpenClaw.He explained, “Loops are now a key part of how we get AI agents to iterate at length to build software.” He added that these loops influence not only how software is built but also how developers decide what to create. In simple words, it is the practice of designing an automated system where an AI keeps working, checking its own results and fixing its own mistakes until a goal is reached, without any human involvement to guide it step-by-step.

Ng outlined three crucial loops that he believes are shaping modern software development: the agentic loop, the developer feedback loop and the external feedback loop. The first loop focuses on AI doing much of the engineering work on its own. As per the tech mogul, developers can provide a product specification and, if needed, evaluation tests. The AI agent can then write code, test its own work and continue improving it until it meets the required standard. He said, “Given a product specification... we can have an AI agent write code, test its work, and keep iterating until the code is bug-free and meets its specification.”The Coursera co-founder described this as a big shift that allows coding agents to work productively for much longer without human support. He cited a personal experience and mentioned that he recently built a typing practice app for his daughter using an AI coding agent. He wrote, "My coding agent could easily work for around an hour, using a web browser to check what it had built multiple times before getting back to me, without needing my intervention."

Humans Still Guide The Product

Even though AI agents have come much better at writing and testing codes, Ng believes that developers now spend less time fixing bugs and more time making important product designs. He said developers constantly review what AI produces and adjust the product vision. While building the typing app, he changed his mind several times about the visual design, the rewards children could unlock and how parents would manage the learning experience. He explained, "For pretty much all the products I'm involved in, I see humans as having a significant context advantage over current AI systems—we know a lot more than the AI system about the users and the context the product has to operate in—and thus humans play a critical role. So long as the human knows something the AI does not, human-in-the-loop is needed to inject that knowledge into the system.”For Ng, the future of software development is not about humans competing with AI. Instead, it is about people and intelligent agents working together in continuous loops, with AI handling repetitive engineering tasks while humans provide the creativity, context and vision that ultimately shape great products.