DeepSeek kernel engineer's poignant essay on AI making his own craft obsolete
teortaxesTex · x · 2026-09-15
DeepSeek kernel engineer Shengyu Liu (interestingLSY) published a striking blog post, translated by teortaxesTex: from ChatGPT's early thousand-token stumbles to reasoning models (o1, DeepSeek R1, Kimi K1.5 Thinking) took two years; to today's tool-wielding agents took another 1.5 years. Part one is his personal struggle as his work renders his beloved craft obsolete; part two draws conclusions the translator 'agrees with'—an unusually candid admission from a frontier-lab systems engineer.
Related event: DeepSeek kernel engineer writes how AI is making his craft obsolete(2 posts)→
More from AGI Musings
- 2024 survey: 10% AI extinction risk is normal among 1,580 AI researchers — NathanpmYoung · 2026-09-15
- Speaker Johnson says tech firms, not Congress, own AI safety — and three CEOs may hold more power than regulators — JOBhakdi · 2026-09-15
- Columbia CS professors dissect the AI panic, urge calm and transparency — vishalmisra · 2026-09-15
- Dean Ball: next AI breakthroughs will come from safety and alignment, not despite it — deanwball · 2026-09-15
- Digital fruit fly brain completes open-world navigation; zebrafish connectome mapping underway at 200TB — udmrzn · 2026-09-15
- Nate Sparrow vs critics: do LessWrong posts and non-peer-reviewed papers prove AI risk? — emax · 2026-09-15