VC Rohit Krishnan's AI takes: GLM and DeepSeek read better than US models, SWE jobs stay
zacharynado · x · 2026-10-24
Zachary Nado retweets Rohit Krishnan's two sets of AI takes.
Recent takes:
- GLM and DeepSeek outputs are currently more pleasant to read than any American model
- The software engineer job market will be the same order of magnitude in five years; "box checker" jobs in coding, accounting and law are really human interfaces to other humans and won't disappear
- AI keeps getting better at learning patterns in front of it, but the last mile stays hard (Zeno's paradox)
- Big companies struggle with AI rollout because they cling to deterministic software structures and can't tolerate mistakes
Earlier takes:
- Frontier labs will keep expanding their inference empires, but profits won't be primarily API-driven; they'll need new businesses to monetize indirectly
- Judgement and taste will prove far harder than expected, not solvable generally across time and domains
- No jobsocalypse, but jobs will change drastically as AI rapidly learns parts of them
More from AGI Musings
- Critical vulnerabilities going parabolic: 14,943 published in Sept 2026 alone — luisdans · 2026-10-24
- Willer: as AI speeds up findings, humans need room to process them — RobbWiller · 2026-10-24
- Stanford sociologist: AI-optimized papers shouldn't crowd out speculative writing — RobbWiller · 2026-10-24
- Sociologist warns AI-optimized papers could erase exploratory thinking in social science — RobbWiller · 2026-10-24
- Zvi's 'The Hackening': experts predict AI-driven internet hostility within 6-12 months — TheZvi · 2026-10-24
- Scoble on three decades of fame: it's a terrible metric of success — Scobleizer · 2026-10-24