If coding is solved, what now?: Measuring the sloppiness of code

A developer writing as "doppp" posted an essay on Earendil arguing that if AI has solved the problem of writing code, the real challenge now is measuring and improving code quality. The post, which reached the front page of Hacker News with 70 points and 81 comments, frames code sloppiness—inefficiency, redundancy, and lack of polish—as the next battleground in developer tooling and software engineering practice.

The essay's core argument is that as large language models and code-generation tools make it trivial to produce working code, the bottleneck shifts from creation to quality. The author describes "sloppiness" as the gap between what code does (its functional behavior) and how it does it (its style, efficiency, maintainability, and adherence to best practices). Where earlier generations of developers competed on the ability to write code that worked, today's bottleneck is writing code that scales, integrates, and survives in production. The post suggests that current metrics—lines of code, test coverage, cyclomatic complexity—are blunt tools and proposes that more granular measures of code quality are necessary as automation handles the mechanical task of coding.

The timing reflects a broader industry inflection point. As of 2024–2026, tools like GitHub Copilot, Claude, and other code assistants have normalized AI-generated code in many workflows. The question shifting among practitioners is no longer "can AI write working code?" but "can AI write good code?" and "how do we grade it?" The post taps into a growing conversation in engineering teams about technical debt, maintainability, and whether speed of delivery—a core pitch of AI coding assistants—is worth the potential cost in code quality and long-term maintenance burden.

Responses on Hacker News suggest the community is divided. Some commenters argue that code quality has always mattered and that the framing of "sloppiness" as a new frontier oversimplifies the discipline of software engineering. Others point out that measuring quality itself is subjective and context-dependent—what is sloppy for a startup's prototype may be appropriate for a regulated financial system. A few responses note that the real issue is not measurement but incentives: developers and teams are rewarded for speed and feature delivery, not for code elegance, so even if sloppiness could be quantified, there's little pressure to reduce it.

The essay arrives as organizations increasingly adopt AI-assisted development. Whether code sloppiness becomes a tracked metric in CI/CD pipelines, or remains a matter of engineering culture and code review, the underlying tension—speed versus quality—remains unresolved. The post does not propose a company or product but instead offers a conceptual framework that toolmakers, teams, and organizations will likely grapple with as AI coding tools mature and the competitive advantage shifts from writing code quickly to writing code well.