Building AI-Intensive Software with AI: Early Results and a Cautionary Tale on Measuring Development Cost
Source
arxiv.org
Author
Victor Barros de Miranda Neves, Kiev Santos da Gama, Vinicius Cardoso Garcia
Date
Why it matters
AI cost-savings ratios are easy to construct wrongly: pricing tokenThe chunk of text a model reads and writes in — roughly three-quarters of a word — and the unit AI usage is billed in.Full definition → under a flat subscription and mispricing the human counterfactual doubled this team's claimed 19.4x down to 9.9x on recheck.
Terms in this piece · Glossary
RAG — Retrieval-augmented generation — fetching relevant documents first and pasting them into the model's context so it answers from your data instead of memory.
token — The chunk of text a model reads and writes in — roughly three-quarters of a word — and the unit AI usage is billed in.
articleThe Psychological Costs of Artificial Intelligence Adoption in Software EngineeringAdam Alami, Elda Paja, Abhishek Tiwari
articleHelpful but Fallible: Developer Experiences of AI Tools Under a Coordinated Industrial Roll-outAndreas Bexell, Rushali Gupta, Lo Gullstrand Heander, Emma S\"oderberg, Per Runeson, Sigrid Eldh, Wei Shi, Konstantin Malysh
articleA Few Pages of Markdown: Committed AI Configuration and Lower Quality Cost after Coding-Agent AdoptionYegor Denisov-Blanch, Shyam Agarwal, Pavel Azaletskiy, Hao He, Rylan Schaeffer, Brando Miranda, Bogdan Vasilescu, Sanmi Koyejo