The Customer Is the Builder: How AI Coding Agents Are Rewriting the Economics of Software
Source
arxiv.org
Author
Ripon Chandra Malo, Zimeng Yu, Tong Qiu
Date
Why it matters
AI agentAn AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.Full definition → task-horizon growth is shifting buy-versus-build decisions for software. The paper defines a substitution frontier and weighs early market evidence, which is mixed, so builders can judge where in-house agent-built tools may beat purchased ones.
Key takeaways · AI-distilled
The authors define a software substitution frontier: the functions customers could build and maintain at an acceptable standard for less than buying them, focusing on organizations whose domain knowledge lets them specify and evaluate their own tools.
Since 2023, the length of benchmarkA standard public test set for comparing AI models — the shared scoreboards behind every "model X beats model Y" claim.Full definition → tasks leading agents complete at 50% success, measured in human working time, has doubled about every 129 days. The authors stress that projections from this trend are an extrapolation, not a forecast of customer switching.
A broad search found 575 potential open-source alternatives to selected commercial products, though some matches still need validation. The Craft suite produced 12 application repositories in 12.0 days, but they remain early versions of unverified completeness and reliability.
Market evidence is mixed: purchased software gained US investment share and tracked installations grew, yet 32% of respondents to an organizational survey reported forgoing a purchase because they could build it with agents. Self-reports do not establish displaced spending.
Terms in this piece · Glossary
AI agent — An AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.
benchmark — A standard public test set for comparing AI models — the shared scoreboards behind every "model X beats model Y" claim.
articleBuilding AI-Intensive Software with AI: Early Results and a Cautionary Tale on Measuring Development CostVictor Barros de Miranda Neves, Kiev Santos da Gama, Vinicius Cardoso Garcia
videoHow AI Coding Agents Will Change Your JobY Combinator
articleThe Devil Is in the Interface: Evaluating How Tool Architecture Shapes Coding Agent BehaviorXiangzhe Xu, Hamidreza Saghir, Qianhui Wu, Marc-Alexandre C\^ot\'e, Tong Wang, Kiran Lakkaraju, Kexin Pei, Xiangyu Zhang