A reliability engineer makes the case that coding is not solved

Ewerlöf, who says he has built his own AI coding harness and used LLM tools in production for four years, argues that code creation is cheap, but that the real cost of software, meaning maintenance, reliability, security, and scale, is exactly where current models still fail. He proposes a three-part test for who owns AI-shipped code: knowledge (do you understand the problem and the tool's limits), mandate (were you actually authorized to make the call), and accountability (are you the one on call when it breaks). The piece names Shopify CEO Tobi Lütke's new term for the failure mode, "slop grenades" (via Business Insider), and quotes a reader comparing Ewerlöf's position directly to DHH's Rails World "Pencils down" keynote, covered on this feed September 25. Someone else submitted the essay to Hacker News two days after it published (the author notes the submission inside the post itself); as of this edition, about 3.2 hours after that submission, the discussion stood at 136 points and 122 comments.

Why it matters: This directly rebuts the optimism this feed has tracked from Huntley's "your product is the factory" line and Matt Pocock's "Software Factory" dictionary entry. The knowledge, mandate, and accountability test gives teams a concrete question to ask before handing a coding agent more autonomy: who is on call when it breaks.