Metraipolis › archive › dc_2b9d0f2a3bcb
Architecture as Capability Equalizer for Coding Agents — arXiv:2608.21747
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2608.21747
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-18T07:42:23.430Z
- SHA-256 of the archived text
daf0228ef892860cb4dc55ed85bd53e64abdb0e891bf57c586dcbaa21a3000f5- Stable URL of this page
https://metraipolis.com/documento/dc_2b9d0f2a3bcb
How to cite
Architecture as Capability Equalizer for Coding Agents — arXiv:2608.21747 — arXiv (en), https://arxiv.org/abs/2608.21747 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_2b9d0f2a3bcb (Metraipolis archive dc_2b9d0f2a3bcb, sha256 daf0228ef892860cb4dc55ed85bd53e64abdb0e891bf57c586dcbaa21a3000f5)
BibTeX
@misc{dc2b9d0f2a3bcb,
title = {Architecture as Capability Equalizer for Coding Agents — arXiv:2608.21747},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2608.21747},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_2b9d0f2a3bcb; sha256 daf0228ef892860cb4dc55ed85bd53e64abdb0e891bf57c586dcbaa21a3000f5},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2608.21747. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
LLM-based coding agents generate complete software systems from high-level descriptions, yet little is known about how the format of architecture specifications affects the quality of generated code or whether this effect depends on model capability. We present a controlled experiment comparing five informationally equivalent specification formats (informal prose, Mermaid diagrams with constraints and ADRs, OpenAPI, C4/Structurizr DSL, and TypeScript interface contracts with ArchUnit-style rules) across six models from three vendor families (Anthropic Claude, OpenAI GPT, Google Gemini). Across 90 multi-turn agent trials, specification format shows a strong format x model interaction. On the strongest models (Sonnet 4.6, GPT-5), format barely matters (quality spread 0.17-0.92). On weaker models, format produces spreads of 0.83-2.42 points, with code-proximate formats (OpenAPI, TypeScript contracts) recovering most of the capability gap. Mid-tier models can consume more tokens than frontier models for worse output when they enter compilation debugging loops that stronger models avoid. Self-validation rates collapse from 100% (Sonnet) to 0% (Gemini Flash) across the capability spectrum. TypeScript contracts triple API route coverage for the weakest model (33% to 100%). Structured architecture specifications serve as a capability equalizer, with value inversely proportional to model strength and the largest returns for cost-optimized deployments. Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-08-25 in cs.AI). Categorie arXiv: Software Engineering (cs.SE) ; Artificial Intelligence (cs.AI); Computation and Language (cs.CL). Pagina del lavoro: https://arxiv.org/abs/2608.21747 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito). Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina. Letto il 2026-10-18T07:37:52.827Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-08-25?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).