HypothesisVery high confidenceCurrent
AI-assisted FDE can create leverage, but local code-generation speed is not a reliable proxy for end-to-end delivery improvement.
Randomize matched delivery tasks and measure elapsed time, review cost, defects, stability and accepted customer outcome.
Valid from2026-08-23
Last reviewed2026-08-23
Review again2026-11-21
Dataheart stanceexploring
Evidence for
DORA 2025 found positive throughput relationships in stronger organizational systems.
DORA — 2025 State of AI-Assisted Software Development · cloud.google.com ↗METR — Early-2025 developer productivity randomized trial · metr.org ↗METR — Updated experiment-design and selection-bias analysis · metr.org ↗METR's later study produced signals consistent with speedup for some selected participants.
DORA — 2025 State of AI-Assisted Software Development · cloud.google.com ↗METR — Early-2025 developer productivity randomized trial · metr.org ↗METR — Updated experiment-design and selection-bias analysis · metr.org ↗Evidence against
DORA continued to observe a negative stability relationship.
DORA — 2025 State of AI-Assisted Software Development · cloud.google.com ↗METR — Early-2025 developer productivity randomized trial · metr.org ↗METR — Updated experiment-design and selection-bias analysis · metr.org ↗METR's early-2025 randomized trial found a 19 percent slowdown.
DORA — 2025 State of AI-Assisted Software Development · cloud.google.com ↗METR — Early-2025 developer productivity randomized trial · metr.org ↗METR — Updated experiment-design and selection-bias analysis · metr.org ↗METR judged its later central estimate unreliable because of selection and measurement effects.
DORA — 2025 State of AI-Assisted Software Development · cloud.google.com ↗METR — Early-2025 developer productivity randomized trial · metr.org ↗METR — Updated experiment-design and selection-bias analysis · metr.org ↗How this claim can fall
Replace this placeholder during human review with an observable result that would materially weaken or falsify the claim.
A published claim remains useful only while it remains possible to prove it wrong.