Public transcript

FULL INTERVIEW: Alex Karp on AI, Job Loss, and the Future of Work

By: TBPN
Length: 28 min
Language: English
Text from: the video's captions
Transcribed: 13 October 2026

Read the full transcript Open on YouTube

Notes in English.

Summary

Alex Karp, CEO of Palantir, sits down with TBBN for a wide-ranging interview covering AI-driven job loss, the future of work, Palantir's hybrid software-and-people business model, and US-China competition. He argues AI will take white-collar jobs, warns of a political backlash that could include nationalizing tech companies, and calls for vocational education reform, new aptitude testing, and honest career guidance. He also defends Palantir's work with the military and says the US must win the AI race against China.

Key points

Karp says there are two ways to have a future: vocational training or being neurodivergent.
He says low-end coding, lawyering, reading and writing are no longer valuable skills.
Karp says AI will likely take white-collar jobs and warns of a political backlash, including possible nationalization of tech companies.
He says Palantir is a hybrid of humans, FDEs, AI FDEs, orchestration and primitives, not a pure software company.
He argues models are 160 IQ against one test but near zero by the 50th step of a thousand institutional tests.
He says constitutional rights should not be extended to adversaries like Iran, and the Department of War must arbitrate battlefield AI deployment.
He calls for vocational education reform, new aptitude testing, and honest warnings about careers and university debt.
He says the US and China are the only two relevant AI powers and the US should win.

Questions it answers

00:00Opening banter and dead hangs

What health metrics does Karp track and how is his dead hang going?

Karp says his dead hang is now 50-55 seconds, up from four minutes discussed last time (as the hosts frame it). He argues dead hang, farmers walk, body weight and VO2 are the few accurate proxy indicators for health.

Karp says his dead hang is now 50-55 seconds.
He calls dead hang, farmers walk, body weight and VO2 the proxy indicators that count for health.

02:02Who has a future in the AI economy

Who will still have a future as AI takes over routine cognitive work?

Karp says there are two ways to have a future: vocational training or being neurodivergent, broadly defined. Skills like low-end coding, lawyering, reading and writing that used to be precious are no longer valuable, so people need to become more like artists and build something unique.

Karp says vocational training or neurodivergence are the two ways to have a future.
He says low-end coding, lawyering, reading and writing are no longer valuable.
He says people with 'normal shape skills' now need to be more like artists.

06:13Palantir's hybrid model and the SaaS apocalypse

How does Karp respond to the idea that AI will replatform enterprise SaaS and to old doubts about Palantir being a services company?

Karp says AI makes it harder to lie about products that don't create value, so software companies that only sold a feeling will be exposed. He says Palantir was never pure software but a hybrid of humans, forward-deployed engineers (FDEs), AI FDEs, orchestration and primitives, and that the services component is now crucial as competitors try to copy it.

Karp says Palantir is a hybrid of humans, AI forward-deployed engineers, orchestration and primitives.
He says AI makes it harder to lie about products that don't create value.
Clients care that Palantir transformed their business in 3 months instead of 3 years.

10:05Why 160 IQ models aren't everywhere

Why aren't genius-level AI models deployed everywhere in institutions?

Karp says the models are 160 IQ against a single test, but institutions require passing a thousand differentiated tests over time, so by the 50th step the effective IQ is near zero. He says clients keep asking for more capacity and invites aligned, technical people to apply to Palantir.

Karp says models are 160 IQ against one test but near zero by the 50th step of a thousand.
He says clients are asking for more capacity, not fewer people.
He invites aligned, technical or smart people to apply to Palantir.

11:30Rights, war fighting, and US-China

How does Karp split domestic rights from war-fighting uses of AI?

Karp says Americans have god-given rights like free expression, the second amendment and privacy that should not be extended to adversaries such as Iran. In the war-fighting context the Department of War must arbitrate deployment, and with only the US and China as relevant AI powers, the US should win.

Karp says constitutional rights should not be extended to adversaries like Iran.
He says there are two relevant powers, the US and China, and the US should win.
He says battlefield data aggregation is lethal and morally justified.

12:53Job loss, nationalization risk, and policy

What does Karp think happens politically if AI takes white-collar jobs?

Karp says AI will likely take jobs, especially white-collar ones, and that a strong movement to nationalize tech companies will arise quickly if nothing is done. He warns the American version of backlash may not be the protective German model but 'hang the rich' without helping the poor.

Karp says AI will likely take white-collar jobs, especially for non-neurodivergent workers.
He predicts a strong movement to nationalize tech companies if reforms don't happen.
He says the American version of backlash could be 'hang the rich' rather than protecting workers.

19:06Education reform and the good outcome

What reforms does Karp propose for education and what does a good outcome look like?

Karp calls for adopting Germany's vocational high school model, changing aptitude testing to capture different forms of intelligence, honest warnings about careers that won't exist, and warning labels on university debt. The good outcome is showing adversaries the US has the best military, AI-enhancing hospitals and manufacturing, and reskilling workers.

Karp praises Germany's three-track high school system with two vocational tracks.
He says US tests are built around industrial-revolution skills and miss neurodivergent talent.
He says the good outcome includes rebuilding manufacturing and AI-enhancing hospitals and infrastructure.

24:00Being misunderstood and the viral question

How does Karp feel about people misunderstanding Palantir's business?

Karp says his value converges with people's inability to understand the business, which makes it nearly impossible to copy, like French luxury or Swiss watches. He finds it frustrating that Palantir's results, including a rule of 127, aren't acknowledged, but views the misunderstanding as a feature.

Karp says Palantir's value converges with people's inability to understand it.
He cites a 'rule of 127' versus the typical benchmark of 40.
Asked $10 million or ChatGPT access in 2012, he jokes he'd take his grad school social life.

Notes

Who has a future in the AI economy

Karp says there are two ways to have a future: vocational training or being neurodivergent (broadly defined).
Skills that used to be precious — low-end coding, low-end lawyering, low-end reading and writing — are no longer valuable.
People with "normal shape skills" now need to become more like artists: look at things differently and build something unique.

Palantir's business model

Palantir was never pure software: it is a hybrid of humans, forward-deployed engineers (FDEs), AI FDEs, orchestration, and "primitives" — tribal knowledge coded into logic and extended in LLMs.
Karp says the services component is crucial, and competitors who mocked it are now trying to hire FDEs.
Clients care about one thing: "you transformed my business in 3 months, it would have taken 3 years."
AI makes it harder to lie about products that don't create value, which threatens legacy SaaS.

Why AI isn't everywhere yet

Models are "160 IQ against a test," but institutions require passing a thousand differentiated tests over time — "by the 50th step it's zero IQ."
Clients are asking for more capacity, not less; Karp invites aligned, technical people to apply.

Politics, rights, and war

Constitutional rights (free expression, second amendment, fourth amendment) should not be extended to adversaries like Iran.
There are two relevant AI powers — the US and China — and Karp says the US should win.
He predicts a strong movement to nationalize tech companies if job losses aren't addressed; the American backlash could be "hang the rich" rather than the protective German model.

Reforms for a good outcome

Adopt Germany's vocational high school model; change aptitude testing to capture different forms of intelligence.
Be honest about which careers won't exist; warn students about university debt.
Rebuild manufacturing, AI-enhance hospitals and infrastructure, and reskill workers.

Text from the video's own captions. Summary and notes written by AI from the transcript, so check anything important against the recording.