Type: url
Original: https://tonybai.com/2026/02/28/agentic-software-engineering/
Status: generated
Summary
The opening instalment of a Chinese-language micro-column on software engineering in the age of AI agents, arguing that the industry is treating coding agents the way no one would treat a brilliant but amnesiac junior developer — letting them push to production without the review discipline their unreliability calls for. Its central claim is that vibe coding buys speed while accumulating trust debt: code generation is no longer the bottleneck, human attention and review bandwidth are, so reading a pull-request diff by eye collapses once an agent emits hundreds of lines spanning several microservices in seconds. The author’s answer, credited to the influence of Ahmed E. Hassan’s work on SE 3.0 and agentic software engineering while arguing that a gap remains between understanding the theory and making it a team’s daily practice, is to put non-deterministic magic inside a deterministic engineering cage, and the post lists the column’s fourteen lectures across cognitive reframing, human-agent collaboration patterns, reliability engineering and platform/team scaling.