This article strikes at the heart of the weirdness in the software industry right now. Programmers making $200K+ are no longer writing code but instead simply supervising Claude Code which writes the code for them.
But instead of being afraid for their jobs, they are excited by how much more productive they are.
Large language model (LLM)-powered agents have demonstrated strong capabilities in automating software engineering tasks such as static bug fixing, as evidenced by benchmarks like SWE-bench. However, in the real world, the development of mature software is typically predicated on complex requirement changes and long-term feature iterations -- a process that static, one-shot repair paradigms fail to capture. To bridge this gap, we propose \textbf{SWE-CI}, the first repository-level benchmark built upon the Continuous Integration loop, aiming to shift the evaluation paradigm for code generation from static, short-term \textit{functional correctness} toward dynamic, long-term \textit{maintainability}. The benchmark comprises 100 tasks, each corresponding on average to an evolution history spanning 233 days and 71 consecutive commits in a real-world code repository. SWE-CI requires agents to systematically resolve these tasks through dozens of rounds of analysis and coding iterations. SWE-CI provides valuable insights into how well agents can sustain code quality throughout long-term evolution.
@carnage4life another part of the problem is the person writing the prompts. There are those who use AI to codegen because they have no experience, and there are those who use it because it’s essentially a keystroke and therefore time saver to some degree.
The latter case will be more likely to correct their AI response when they’re architecturally wrong “long term”. You HAVE to be like Fletcher from Whiplash if you don’t want this codebase you’re using AI on to quickly turn to a ball of mud.
@carnage4life @cate a very weird take, like a glimpse into another reality. “You might imagine this would unsettle and demoralize programmers. Some of them, certainly. But I spoke to scores of developers this past fall and winter, and most were weirdly jazzed about their new powers.”
yeah because a decent number of us who are “unsettled and demoralized” (and got laid off thanks to AI boosterism) are simply… not around to be interviewed about workflows anymore. survivorship bias at its best 🤷🏻
Nico
Dare Obasanjo
Jacques Chester
haifisch
Armando
Kat