Synthesis

Slow adoption: how fast humans really take up technology

Dave's thesis, echoed by others, that real adoption of AI runs far slower than hype because of price, user experience and habit, so debates should be grounded in usage data.

Type
Synthesis
Confidence
Medium
Sources
11 videos
First published
Updated

Across 18 episodes the recurring claim is that adoption lags capability. Dave coins ‘humans don’t metabolize things very fast’ and says general-purpose technologies take about twenty years to deliver the hyped use cases. He compares AI today to the 1982 Macintosh and the 1997 web, and notes his first Macintosh cost about $6,000 in today’s money. He grounds this in numbers: developers in single-digit or tens of millions against billions of software users, and ChatGPT near 800 million users. Others put it differently: Amir Efrati (The Information) says AI has a discovery problem, not a demand problem, at roughly 1% penetration; Jess says economics, especially price, is the best lens, and that the tech industry wrongly assumes everyone wants efficiency. Brit calls the present phase evolutionary, not revolutionary, and Jess and Brit expect much AI value to become invisible. Figures are inconsistent between speakers (ChatGPT users 800 million versus about a billion; 0.4% and 1% adoption) and rarely sourced. This is largely one thesis voiced by Dave, so treat the recurrence as one show’s view.

Key points

  • Dave: humans metabolise technology slowly; do not extrapolate hype into near-term enterprise adoption — 5 episodes. (Dave · Stripe Buys OpenRouter, OpenAI (Aug 2026) ▶ 48:08)
  • Ground the debate in adoption numbers: usage is small against the addressable population — 5 episodes. (Dave · Why the Met Gala Hates Tech, E (May 2026) ▶ 30:13) (Amir Efrati · Amir Efrati (May 2026) ▶ 36:43) (Adrian · Adrian (Oct 2026) ▶ 29:20)
  • Price, user experience and ordinary choices drive adoption curves, not capability or efficiency — 3 episodes. (Dave · Amir Efrati (May 2026) ▶ 18:21) (Amir Efrati · Amir Efrati (May 2026) ▶ 29:18) (Adrian · Adrian (Oct 2026) ▶ 39:52)
  • Much AI will become invisible, running in the background of products people already use — 3 episodes. (Brit · Amir Efrati (May 2026) ▶ 28:25)
  • Current AI tools are partly toys or incremental: lots built, few kept; phase is ‘evolutionary, linear’ — 2 episodes. (Sam · SaaS Companies Beware: AI Is T (Jan 2026) ▶ 51:07)
  • Historic analogies for pace: early PC (30 years to cheap), web in 1997, Macintosh not yet invented for agents — 4 episodes. (Dave · What Tim Cook’s “Executive Cha (Apr 2026) ▶ 51:23)

Where guests disagree

Is slow uptake mainly about price, or about people’s own choices? ⚠️

  • Dave: the only explanation he sees for flattening is price, since frontier models are too expensive (Dave · Amir Efrati (May 2026) ▶ 18:21)
  • Amir and Jess: people do not simply adopt the efficient option; AI has a discovery problem and varied personal choices (Amir Efrati · Amir Efrati (May 2026) ▶ 29:18) (Jess · Amir Efrati (May 2026) ▶ 28:45)

How many people use ChatGPT? ⚠️

  • Dave: still not beyond 800 million and growth is declining (other hosts dispute this) (Dave · Why the Met Gala Hates Tech, E (May 2026) ▶ 22:16)
  • Adrian (guest): about a billion use it and roughly ‘billion5’ have tried and not stuck (Adrian · Adrian (Oct 2026) ▶ 29:20)

Stories & examples

  • Dave compares AI tools to the early personal computer: his Macintosh cost $6,000 in today’s dollars and it took 30 years to make computers cheap and accessible; he says we are in the second phase of that story. (Dave · What Tim Cook’s “Executive Cha (Apr 2026) ▶ 51:23)
  • Sam says AI tools are like the fart-app era of the iPhone: lots of toys get built and most go away. (Sam · SaaS Companies Beware: AI Is T (Jan 2026) ▶ 51:07)

Numbers

  • Codex users (Dave: last number): 4000000 people (host’s framing) (Dave · Why the Met Gala Hates Tech, E (May 2026) ▶ 30:19)
  • ChatGPT users ceiling cited by Dave: 800000000 people (host’s framing, approx.) (Dave · Why the Met Gala Hates Tech, E (May 2026) ▶ 22:16)
  • Amir’s estimate of AI market penetration: 1 percent (self-reported, approx.) (Amir Efrati · Amir Efrati (May 2026) ▶ 36:43)
  • AI adoption in August: 0.4 percent (self-reported) (one of the hosts · iPhone Duo, Meta Muse & Why AI (Sep 2026) ▶ 31:02)
  • ChatGPT weekly or regular users, as Adrian recalls: 1000000000 users (self-reported, approx.) (Adrian · Adrian (Oct 2026) ▶ 29:22)
  • Cost of Dave’s first Macintosh in today’s dollars: 6000 dollars (host’s framing, approx.) (Dave · What Tim Cook’s “Executive Cha (Apr 2026) ▶ 51:23)

Caveats & warnings

  • Dave says charts showing Claude Code and Codex usage flatlining may not be true. (Dave · Amir Efrati (May 2026) ▶ 18:10)
  • Jess says she is not anti-AI and expects AI to transform society; her point is about uneven adoption. (Jess · Amir Efrati (May 2026) ▶ 23:13)
  • Sam admits he builds lots of tools and uses almost none, and is not convinced AI will change his app stack. (Sam · SaaS Companies Beware: AI Is T (Jan 2026) ▶ 49:42)
  • Dave concedes AI tools are still complicated and hard to use, and that you have to put a room together to teach people. (Dave · What Tim Cook’s “Executive Cha (Apr 2026) ▶ 51:41)
  • Dave argues that almost no one is using AI engineering tools in the grand context of the world: the developer population is around 10 million and Codex’s last number was 4 million. (Dave · Why the Met Gala Hates Tech, E (May 2026) ▶ 30:13)

How much to trust this

  • Evidence: 18 episodes, 5 independent speakers (the four hosts and unidentified host speech count as one voice, the show; guests counted individually).

Sources

  1. (01)

    Why the Met Gala Hates Tech, Elon vs OpenAI Drama, and the Rise of Chinese AI Models

    YouTube

    22:16 into “Why the Met Gala Hates Tech, Elon vs OpenAI Drama, and the Rise of Chinese AI Models”

  2. (02)

    SaaS Companies Beware: AI Is The New UI (Anthropic's Claude Code and Cowork)

    YouTube

    49:42 into “SaaS Companies Beware: AI Is The New UI (Anthropic's Claude Code and Cowork)”

  3. (03)

    iPhone Duo, Meta Muse & Why AI Finally Scared Dave Morin

    YouTube

    31:02 into “iPhone Duo, Meta Muse & Why AI Finally Scared Dave Morin”

  4. (04)

    Normal People Don't Want AI Agents (And Silicon Valley Can't Accept It)

    YouTube

    18:10 into “Normal People Don't Want AI Agents (And Silicon Valley Can't Accept It)”

  5. (05)

    Stripe Buys OpenRouter, OpenAI vs. Anthropic & the AI Backlash

    YouTube

    48:08 into “Stripe Buys OpenRouter, OpenAI vs. Anthropic & the AI Backlash”

  6. (06)

    Google Just Won AI, TikTok’s “Self-Control Badges”, ChatGPT Enters Group Chats

    YouTube

    33:22 into “Google Just Won AI, TikTok’s “Self-Control Badges”, ChatGPT Enters Group Chats”

  7. (07)

    What Tim Cook’s “Executive Chairman” Move Really Means | Musk vs OpenAI, xAI and Cursor Rumors

    YouTube

    51:23 into “What Tim Cook’s “Executive Chairman” Move Really Means | Musk vs OpenAI, xAI and Cursor Rumors”

  8. (08)

    Why SpaceX's $1.75T IPO Makes No Financial Sense (And Doesn't Have To)

    YouTube

    16:55 into “Why SpaceX's $1.75T IPO Makes No Financial Sense (And Doesn't Have To)”

  9. (09)

    Meta Muse vs Instinct vs OpenAI Dots: Do AI Agents Have Any Moat?

    YouTube

    29:20 into “Meta Muse vs Instinct vs OpenAI Dots: Do AI Agents Have Any Moat?”

  10. (10)

    AI Frontier Lab Is Essentially Doomed | Open AI Offered U.S. Equity, Xbox Is In Trouble, Post-AGI

    YouTube

    48:02 into “AI Frontier Lab Is Essentially Doomed | Open AI Offered U.S. Equity, Xbox Is In Trouble, Post-AGI”

  11. (11)

    Every CTO Is Quietly Switching Off Anthropic and OpenAI | GLM 5.2, Vox Americanus, Claude Fable 5

    YouTube

    34:23 into “Every CTO Is Quietly Switching Off Anthropic and OpenAI | GLM 5.2, Vox Americanus, Claude Fable 5”

(By email)

Compounding Insights