The Irreducible Human: Six Things We Still Do That Machines Can’t

Source: Nate’s Newsletter Date: 2025-05-27 URL: https://natesnewsletter.substack.com/p/the-irreducible-human-six-things

Summary

Grounded in Polanyi’s Paradox (“we know more than we can tell”), Nate argues that six fundamental human capabilities with significant economic value cannot be automated by LLMs because they rely on tacit knowledge that cannot be articulated or “tokenized.” Written in May 2025 as a counterweight to Google, Anthropic, OpenAI, Microsoft, and Nvidia announcements — and explicitly pushes back on Bill Gates’ claim that AI will handle “most things.”

Implications

AI economics thread. The Polanyi’s Paradox framing is the most intellectually rigorous version of the “irreducible human” argument. If tacit knowledge is genuinely un-tokenizable, the economic moat for human labor isn’t speed, creativity, or even judgment — it’s the categories of knowledge that resist explicit articulation entirely. That narrows the field considerably.

Enterprise adoption thread. Nate’s willingness to disagree with Gates’ maximalist AI displacement thesis is a calibration signal for his practitioner audience: not everything will be automated, the specific domains matter, and overstating AI capability leads to bad workforce decisions in both directions (over-investing in replacement and under-investing in AI-augmented versions of irreducible work).

Watch: Whether the six domains Nate identifies remain automatable boundaries through 2026-2027 — if any of them get breached by capability advances, the tacit knowledge argument loses a pillar and the displacement projection changes significantly.