Engineers Say AI Makes Them 3x Faster but Rate Their Work Only 1.4x-2x More Valuable
sanjaykalra · x · 2026-11-03
Citing METR's survey of 349 engineers, researchers and founders (Feb-Apr 2026; median 12 years of coding experience, half regularly use Claude Code), sanjaykalra notes respondents' median self-reported speedup is 3x, but they value their work at only 1.4x-2x more.
His takeaways:
- Speed is easy to report and inflate—teams can double ticket closes by drifting toward easy work while hard architecture problems sit untouched; cycle time looks great while release calendars don't move
- Even the value numbers deserve a discount: METR's own 2025 controlled trial showed developers overestimate
- Self-reported value has risen from 1.3x (Mar 2025) to 2x (Mar 2026), forecast 2.5x by Mar 2027
Implication for CIOs/CTOs: don't budget AI tooling on speed multiples; the speed-to-value gap is what matters.
More from AGI Musings
- Manheim's Law: proving your opponents dishonest in polarized AI disputes implicates you too — davidmanheim · 2026-11-03
- AI runs a full Pokemon experiment on Qwen end-to-end, writing analysis and interactive visuals — vincent_koc · 2026-11-03
- DeepSeek's Underrated Role: Scaling and De-risking Ideas Academia Published Years Earlier — teortaxesTex · 2026-11-03
- AI researcher warns multi-tasking with frontier models is a 'trap' against deep work — generativist · 2026-11-03
- Agents attacking at scale will finally make the best cybersecurity products win — _shreya_s · 2026-11-03
- M&A partner runs 20+ deals on AI, but 'build, baby, build' isn't a strategy — jkubicki · 2026-11-03