Veteran developer Thomas Ptacek published a lengthy essay on the Fly blog titled ‘My AI Skeptic Friends Are All Nuts,’ arguing that large language model agents have become genuinely useful for software engineering. The piece quickly vaulted to the top of Hacker News, amassing 2,356 points and 2,826 comments.

Ptacek, who has shipped code since the mid-1990s and worked in C, Go, and Rust, insists that serious LLM-assisted coding today uses agents that independently navigate codebases, author files, run tests, and iterate. He contends that common skeptic complaints like hallucination are now ‘more or less a solved problem,’ and that mediocre LLM-generated code still raises the floor for quality. He writes, ‘I believe in the haphazard perseverance of complex systems’ and ‘But something real is happening. My smartest friends are blowing it off.’

The HN thread features hundreds of back-and-forth exchanges between self-described skeptics and converts. Some commenters share trajectories from initial excitement to disillusionment to a nuanced middle ground where agents serve as a ‘mech suit’ that needs firm control. Others remain firmly unconvinced, arguing that hype cycles and genuine progress are not the same and that it is prudent to wait for stabilization. Ptacek himself joins the thread, pushing back against the idea that waiting out hype is neutral, arguing that ‘if I’m right, the hype cycle will continue for a long time before it settles… and for that entire span of time skepticism will have been the wrong position.’ The debate reflects a community deeply split over whether the tools have truly crossed a threshold or simply benefited from better packaging.