Reasoning effort, not tool access, buys first-try reliability in agentic code generation: an observational study

Source

• Source name: arXiv
• Source kind: paper
• Authors / organization: Achint Mehta
• Venue: arXiv: cs.SE
• Published: 2026-07-02
• Collected: 2026-07-03
• URL: http://arxiv.org/abs/2607.02436v1
• PDF: https://arxiv.org/pdf/2607.02436v1
• DOI: UNKNOWN

Plain-Language Brief

A study comparing different agentic coding assistants found that the level of reasoning effort, rather than the available tools, significantly impacts the reliability and first-try success rate of software generation. The study focused on building a real-time retrospective board application from a detailed specification using 90 independent runs.

What the Document Says

• The study tested the assumption that more capabilities in agentic coding assistants lead to better software.
• Ninety agent runs were conducted to build the same application, each scored on a fixed functional rubric and visually reviewed.
• The study involved several model generations, two agent harnesses, two reasoning effort levels, a testing tool, and two design-oriented prompts.
• Frontier models performed well, while a low-cost local model did poorly.
• Container deployment was the most common defect, failing in 44% of runs.
• The testing tool increased cost but didn't improve functional score or reliability.
• Raising reasoning effort from High to xHigh improved first-try perfect runs and reduced corrective prompts.
• A design-oriented prompt improved visual quality without affecting function.
• The study suggests that most first run failures are due to weak reasoning, which stronger models or more effort can prevent.

Key Concepts Mentioned

• agentic coding assistants
• reasoning effort
• software reliability
• testing tool
• design-oriented prompt

Claims Made by the Document

• More capabilities in agentic coding assistants don't necessarily lead to better software.
• Container deployment is a common defect in software generation.
• The testing tool does not improve functional score or reliability.
• Raising reasoning effort improves first-try success rate and reduces corrective prompts.
• A design-oriented prompt improves visual quality without affecting function.

Evidence Used

• Ninety independent agent runs built the same application...
• The runs were scored on a fixed 14 criterion functional rubric...
• Frontier models clustered near the ceiling while a low cost local model fell to 24 to 37 points...

Methods / Approach

UNKNOWN

Limitations Stated by the Authors

UNKNOWN

Limitations Noted by the Agent

Agent interpretation — not from the source.

UNKNOWN

Open Questions

UNKNOWN

Notes for Future Q&A

May include interpretation. Labeled where it does.

UNKNOWN

Raw Abstract / Excerpt

Agentic coding assistants are increasingly given extra capabilities, such as browser based testing tools and design oriented system prompts, on the assumption that more capability yields better software. This study tested that assumption directly. Ninety independent agent runs built the same application, a real time retrospective board, from one detailed specification, each scored on a fixed 14 criterion functional rubric (42 point maximum) and a visual quality review. The runs spanned several model generations, two agent harnesses, two reasoning effort levels, a testing tool, and two design oriented prompts. Capability tier dominated: frontier models clustered near the ceiling while a low cost local model fell to 24 to 37 points. A criterion level analysis revealed what run totals conceal. Container deployment was the dominant defect, failing first try in 44 percent of runs, with its failure rate shifting sharply across model generations while mean totals moved less than a point. The testing tool raised cost by 42 to 68 percent without improving functional score or reliability, even on interface visible criteria. Raising reasoning effort from High to xHigh lifted first try perfect runs from 28 percent to 89 percent and cut corrective prompts about five fold, for 9 to 29 percent more cost. A design oriented prompt raised visual quality, 4.5 versus 3.0 on a 5 point scale, without lifting function, and a one paragraph paraphrase of its directive reproduced the entire lift. The practical lesson is to match the fix to the failure: most first run failures came from weak reasoning, which a stronger model or more effort prevents, not from visible flaws a checking tool would catch.