Alex Ewerlöf · 2026-09-26 · notable
Alex Ewerlöf — coding is not solved, and AI can't be held accountable
Alex Ewerlöf, an engineer who builds his own LLM coding harness, argues that 'coding is solved' ignores maintenance, reliability and security, which are most of the cost. The essay reached 349 points on Hacker News.
An early adopter of AI coding tools explains why writing code got cheap but owning it did not.
What is it?
In 'Coding is NOT solved', Alex Ewerlöf says he is not anti-AI: he adopted LLM coding tools early and built his own harness. His point is that people who call coding solved count only the cost of creating code, while maintenance, reliability, security and scalability make up most of the cost of software in production. He adds that even functional requirements are not a solved problem.
How does it work?
Ewerlöf builds the case in steps. Coding is about logic, and LLMs struggle with logic and with long inputs. Their output is stochastic, so even a model that passes every eval can still be unreliable. And AI cannot be held accountable, because 'the worst thing you can do to AI is to unplug it'. He also points to bugs in Claude Code, noting that Anthropic controls the model, the harness, the prompt and the runtime.
Why does it matter?
Ewerlöf argues that whoever ships code is accountable for it, however it was produced, and that you cannot be accountable for what you do not understand. He predicts many engineers will become technical product managers, AI managers, AI deployment engineers and AI quality engineers rather than disappear, and warns against sharing valuable knowledge with cloud AI providers.
Who is it for?
software engineers and engineering managers