The Second Station: why 5x-ing one factory station doesn't save the line
saranormous · x · 2026-11-10
Sarah Armstrong's essay 'The Second Station' on AI robotics go-to-market: speeding a plant's slowest station from 100 to 500 parts/hour yields only 25% line gain when the next station caps at 125—the plant manager will buy someone else's faster second machine, so robot makers must help fill the whole line or rethink flow. On pharma, AI has accelerated target-to-candidate time, but clinics still consume most of a decade and the money; a better molecule only marginally shifts odds. Core thesis: AI's industrial upside is gated by the slowest station, not single-point capability.
More from AGI Musings
- After ICLR review chaos, researcher calls for centralized AI use in peer review — JessicaHullman · 2026-11-10
- "In AI 3.0 every product is a harness pointed at a database via tools" — rickasaurus · 2026-11-10
- KOL quips the bottleneck may be research staff, amid Anthropic-METR capability assessment rift — teortaxesTex · 2026-11-10
- RBA official: AI investment boom will lift inflation before productivity gains arrive — rohanpaul_ai · 2026-11-10
- Non-native English speaker mocked for using ChatGPT to write Reddit posts — Naive-Consequence110 · 2026-11-10
- iamtrask: agent stability benchmarks measure stability, not intelligence — iamtrask · 2026-11-10