May 6, 2026
Ctrl-Alt-Delusion
Vibe coding and agentic engineering are getting closer than I'd like
Even the careful coders are admitting the robot is driving more than they’d like
TLDR: Simon Willison says even careful programmers are starting to trust AI-written software without checking every line, because the tools are getting that convincing. The comments split fast: some see a productivity miracle, others see a future built on unread machine-made messes and weirdly annoying robot comments.
A respected software veteran just confessed something that hit the comments like a cold splash of water: the line between "I’m carefully using AI" and "the machine is kind of just doing it" is getting blurry. In his podcast chat, Simon Willison said he used to draw a neat boundary. "Vibe coding" was the reckless, fingers-crossed version. "Agentic engineering" was the responsible pro version. Now? Even he’s shipping work without reading every line, because the AI is often just... good enough.
That confession set off a mini morality play in the community. One camp basically yelled, absolutely not. User singpolyma3 said they’re too opinionated to let any code slide if it doesn’t match their vision, human-made or robot-made. Another commenter, QuantumNomad_, went full apocalypse mode, imagining a future buried under billions of lines of unread machine-made spaghetti that nobody can fix. That’s the nightmare fuel.
But there was also a messy middle. Gabriela_c admitted AI sometimes writes a better first draft than she would, then immediately dragged one of the internet’s pet peeves: AI-generated comments. Apparently the real villain may not be the code, but the cheery, soulless little notes sprinkled on top of it. Meanwhile, another commenter zoomed in on Simon’s confidence that a simple task would “just work,” which feels like the exact sentence that summons bugs out of thin air. The mood? Equal parts awe, guilt, and please don’t let the robots leave fake helpful notes in my files.
Key Points
- •Simon Willison says a podcast discussion led him to realize that his previous distinction between vibe coding and agentic engineering is becoming less clear in practice.
- •He describes vibe coding as using AI to generate software without closely reviewing the code, which he considers acceptable for personal tools but not for software used by others.
- •He defines agentic engineering as professional software development that uses AI tools while still applying expertise in security, maintainability, operations, and performance.
- •Willison says AI coding tools have increased the scope of work he can take on, while his goal remains building higher-quality production systems faster rather than lower-quality software more quickly.
- •He notes that increasingly reliable coding agents such as Claude Code have led him to stop reviewing every generated line for some tasks, and he compares that to relying on internal services built by other engineering teams.