A substantial share of builders' voiced design intentions never reaches the typed prompt, leaving design choices to the production stack's defaults¶

Claim · All claims
Evidence · 1 study · qualitative r2 · q2
Currency · newest study 2026 · not yet checked against its sources · not reviewed by a person · recheck by 2029: tool and model accuracy changes quickly

Subclaims¶

q2 i? Across the eight prompt cycles with a usable speech window, builders voiced 38 atomic intents and typed 27, so 71% were transmitted and 29% dropped between mind and keyboard. → Léonard Boussioux* 2026

Evidence¶

Léonard Boussioux* 2026¶

Léonard Boussioux, Ziyi Zhao, and Kanghyun Cho*. (2026). One Tool, One Taste? How Vibe Coding Trades Collective Diversity for Individual Creativity. arXiv. https://arxiv.org/abs/2609.38183 doi:10.48550/arxiv.2609.38183

q2 · i? · qualitative · r2

Process-layer translation-loss analysis comparing atomic intentions voiced in the 240 seconds before submission against typed prompt requirement units, for eight cycles spanning six of the nine builders. The article states the denominator is small, treats the percentage as an illustration, and notes the rate is conditional on verbalized intent.

"builders voiced 38 atomic intents and typed 27 of them, so 71% were transmitted and 29% were dropped between mind and keyboard (Figure 12)."

Discussion¶

Related Claims¶

• Originality is almost never invoked in builders' spoken acceptance verdicts, and defects as well as defaults exit the loop by acceptance without verification — related
• Chat-driven vibe-coding builders converge on a single generate-and-check script whose dominant time cost is dead waiting for the tool — related
• Edited voice prompts did not significantly differ from text prompts in first-attempt success — related
• Students reported mixed, task-dependent perceptions of the efficiency and speed of voice versus text input — related