Vibeleaderboard
← All Intel
Intel / article

The EPIC Framework for Spec-Driven Development

Source
arxiv.org
Author
Bruno Claudino, Ruizhe Xing, Yuanhao Zuo, Tyler Menezes, Anita Sarma, Kostadin Damevski, Zixuan Feng
Date

Context

has a developer write a specification, plan and tasks before a writes code. Templates exist but do not help a developer judge whether a spec is complete or clear. In this paper, submitted October 6, 2026, researchers scored the artifacts of 114 open-source repositories against the ISO/IEC/IEEE 29148 requirements standard and drew practices from the best-scoring ones.

The result, EPIC, has 40 practices across 10 quality dimensions. The paper reports that the majority of 15 practitioners endorsed every practice. Repositories in the top third by spec quality spent 11.8 percent of commits on bug fixes versus 20.4 percent in the bottom third, with a median of four times as many contributors.

That comparison is observational. Better specs may travel with stronger teams, and the abstract does not test whether adopting EPIC reduces bugs.

Key takeaways · AI-distilled
  • One interviewed team writes "must" instead of "should" in instructions because the agent may treat "should" as optional; the authors note agents fill gaps in instructions with their own assumptions.
  • The practices come from scoring the artifacts of 114 open-source SDD repositories against ISO/IEC/IEEE 29148 and distilling what the highest-scoring ones did.
  • A majority of the 15 SDD practitioners consulted endorsed every one of the 40 practices.
  • In the study, repositories in the top third of specification quality spent 11.8% of commits on bug fixes versus 20.4% in the bottom third, and had a median of 4x as many contributors.
Terms in this piece · Glossary
  • spec-driven development — Writing and agreeing the specification first, then having an agent implement against it — the spec, not the code, becomes the artifact you maintain.
  • Agent mode — A mode in which one agent can inspect, edit, run commands, and iterate toward an outcome.
  • Agent — A system in which a model repeatedly chooses operations, a harness carries them out, and the results return to the model to inform its next choice.
Recommended reads
Comments

Checking sign-in…

Loading comments…