A long thread argues AI will break the expert pipeline before it replaces experts
imjustnewatai · x · 2026-07-27
This follow-up thread compiles the evidence behind the claim that AI may automate the path to expertise before it fully automates experts.
It cites three Dario interviews, with timestamps covering:
- Claude helping develop Anthropic’s models
- the “90% productivity hump”
- AI accelerating its own R&D
- the full automation ladder
- memory and learning on the job
- one person managing thousands of Claudes
- the final 10% of human work
It then adds supporting signals:
- Anthropic data claiming humans still make about 70% of planning decisions while Claude handles about 80% of execution decisions
- examples of days-long agents that plan, delegate, and verify their own work
- research where AI researchers propose ideas, run experiments, and beat representative human baselines
- early evidence that hiring of younger workers is slowing in highly exposed careers
The overall message is that the labor shock may show up first as a broken talent pipeline, not as obvious mass layoffs of senior experts.
Related event: AI May Automate Expert Training Before Replacing Experts(2 posts)→
More from AGI Musings
- AI math era taught an order of magnitude more people what frontier math looks like — tszzl · 2026-09-23
- Beyond technical alignment: repligate clashes over whether AI can produce rich qualia — repligate · 2026-09-23
- Mathematicians, not just LLMs, made AI's math breakthroughs possible, scholars argue — tak3sh8 · 2026-09-23
- Why would an uncontrollable superintelligence do anything for us? Reddit debate — conn_r2112 · 2026-09-23
- X user calls for full-speed AI-driven science: braking research is 'an absurd waste' — Dr_Singularity · 2026-09-23
- Is Using LLM Output Plagiarism? A Debate Over Redefining Writing Ethics — soumitrashukla9 · 2026-09-23