[HN]score: 0.35
Faros.ai data across 22K devs suggests LLM use destroys team-level value
May 31, 2026
Faros.ai telemetry data spanning 22,000 developers and 4,000 teams — pulled from Jira, GitHub, and CI/CD pipelines — shows teams using LLMs in software development performing worse on operational metrics than non-LLM teams on average. The author argues current LLM usage patterns are net value-destructive, not productivity-enhancing.
HOW THIS AFFECTS YOU
●
builderWorth scrutinizing your team's actual cycle time and defect rate data before assuming LLM tooling is helping — the aggregate signal here is negative.
●
founderIf this data holds, the productivity narrative underpinning AI coding tools is weaker than marketed, which affects both your own tooling choices and competitive positioning.
●
investorChallenges the core value-creation thesis for AI coding assistant companies; 22K-developer sample from a telemetry vendor is more rigorous than typical survey-based productivity claims.