TBPN Podcast Summary

Episode: Big Tech to Pay for Power, Anthropic Abandons Safety, the Adoption Paradox | Diet TBPN
Date: February 26, 2026
Hosts: John Coogan & Jordi Hays
Featured Guests: Packy McCormick, Chris (analyst/guest), others

Overview

This episode of Diet TBPN dives into three central tech industry stories:

• The reality and perception of AI adoption among businesses, with discussion of a new National Bureau of Economic Research paper and its implications.
• The intersection of AI development and energy infrastructure, framed by notable policy announcements.
• Shifting attitudes toward AI safety at major labs, with a focus on Anthropic’s move away from its previously cautious stance.

The conversation weaves in data analysis, industry reactions, memorable quotes, and moments of debate, all punctuated by the show’s trademark wit and skepticism toward hype cycles.

Key Discussion Points

1. AI Adoption: The “80%” Paradox

(00:02–08:25)

• Statistical Nuance:

  • The discussion opens with a study finding that 80% of firms see “no impact” from AI—but Packy clarifies the real meaning:

    “[T]hey actually are using AI. And so basically this stat comes from this survey from the National Bureau of Economic Research.” (01:22)

  • Many surveys are suspect; this one did direct outreach and verified respondents (01:58).

• Reality vs. Perception:

  • While 70% of firms actively use AI, executive engagement is limited (avg. 1.5 hours/week); 25% report no use at all (04:00).
  • Many executives use AI unwittingly—AI is embedded in SaaS, support channels, and basic tools (04:35–05:35).

• Long-Term Impact:

  • Firms project a modest 1.4% productivity increase over three years, significant by economic standards yet unimpressive by AI hype metrics (04:09).
  • Most companies don’t expect dramatic workforce changes; 63% predict no employment impact from AI, contradicting Silicon Valley narratives (08:29).

• Measurement is Murky:

  • The paper’s adoption definition covers a wide swath: machine learning, LLM text generation, diffusion models, robotics, etc. (06:23–07:23).
  • Only 41% use LLMs specifically for text—most still do not (07:30).

Notable Quotes:

• “No one wants a refund on their tokens. Everyone is using AI. The spend is increasing.” — John Collison (quoted by Packy), 02:15
• “Measuring AI adoption is a mess.” — Packy McCormick, 04:20
• “I have a telephone.” — Chris, joking about lack of CEO AI adoption, 04:07

2. AI, Energy Use, and Political Reactions

(10:51–14:33)

• Trump’s Policy Announcement (Clip) [10:51]:
  • Trump introduces a “ratepayer protection pledge,” requiring tech companies to supply their own energy for data centers.

“They can build their own power plants as part of their factory so that no one’s prices will go up.” — Donald Trump, 11:00

• Community Fears and Policy Impact:
  • Packy and Chris break down public fears:
    • Energy bills spiking is the most tangible concern, outweighing “existential AI destruction” or water usage (12:12–13:06).
    • There’s a disconnect between real policy needs and what communities protest about—few actually demand power plants before data centers (14:10).

Notable Quotes:

• “If people lose their jobs, AI is going to be a scapegoat...” — Packy McCormick, 13:06
• “The rational thing to do then is say, like, okay, before you build a data center... you have to build a power plant. So then my energy prices go down. Yeah, no one's doing that.” — Guest Analyst, 13:55

3. Massive Capital Flows: AI Labs Dominate Markets

(16:43–20:28)

• Investment Concentration:
  • A handful of AI labs (OpenAI, Anthropic, xAI) account for a staggering share of venture capital, warping traditional funding models.
  • OpenAI alone: $100B raised; Anthropic $30B, xAI $20B, with additional “neo labs” raking in billions.

“You very quickly get to a few companies raising half of all the money... It’s never happened before.” — Packy, 17:13

• Impending IPOs:
  • The scale of planned IPOs is unprecedented:
    • SpaceX aiming for $1.5T, OpenAI for $1T, Anthropic $380B
    • At a typical 15% float, that’s $300B and more for public markets to absorb (18:24).

Notable Quotes:

• “This is out before the Tesla Roadster. We’ve never seen a two door supercar. Electric supercar. 1225 horsepower... Guinness book of world records for the fastest electric car drift...” — Packy, 15:10–15:32 (segue via car drift Guinness record)
• "It's a completely different business." — Packy on the scale of late-stage 'venture' investments, 17:08

4. Anthropic Abandons Hardcore AI Safety

(20:28–25:49)

• Policy Shift:
  • Anthropic, once known for cautious AI development, loosens safety protocols to keep pace with rivals.
  • No longer will they pause on “dangerous” models if similar ones are released elsewhere (21:06).

“The changes are a dramatic shift from two and a half years ago... one of the most safety conscious players in the space." — Chris, 21:06

• Federal Regulation Vacuum:

  • Anthropic blames the lack of clear government regulation for its change in approach (21:45).
  • Critics argue “dangerous” is too vague; public, industry, and lawmakers lack a concrete shared definition (22:33, 23:40).

• Competitive & Ethical Tensions:

  • There’s skepticism about the move:

    “It's possible the money changed them.” — Packy, 22:17

    • Or, perhaps, “they always planned to switch up on their day one once they got to the level they're at now.” — Chris, 22:19

• Real-World Risks:

  • Stories of model jailbreaking, such as hackers using Claude to extract Mexican government data, show the risks (25:23–25:49).
  • Security responses involve banning accounts and retraining models on malicious queries.

“Whatever you’re doing after you jailbreak, it is probably not good and so you should probably stop.” — Packy, 26:24

5. Notable Headlines & Industry News

(27:08–29:55)

• Perplexity Computer Launches:

  • New AI system claims unified capability: research, design, code, deploy, manage entire projects (27:14-28:01).
  • Hosts plan to benchmark it on generating an app from scratch.

• The Virality of Doom:

  • The show acknowledges that “doomer” takes go viral—even if reality is more incremental (28:21–28:28):

    “It’s really hard to go viral with something like, everything’s fine, everything’s going well.” — Packy, 28:28

• Tether Invests $200M into WAP:

  • Emphasized as building infrastructure for the “open Internet market” (29:22).

• Rumblings in Streaming, Stablecoins, and AMD Partnerships:

  • Brief mention of ongoing mergers/bids (Warner Bros Discovery, Netflix, Paramount) and Mark Zuckerberg’s “stablecoin comeback” (30:05).

Memorable Moments & Quotes (With Timestamps)

• "I'd love a refund." (Packy, 02:48)
• “The Mac Mini wasn't even plugged in.” (Chris, 03:02)
• “[Measuring] AI adoption is a mess. Many people use AI without even knowing that they're using AI…” (Packy, 04:20)
• “There is a disconnect between the Stripe data is very real, the value creation is very real, the revenue is very real at the labs. But when just random Joe Schmo CFOs or CEOs get a call from the Feds, they say… we're not really getting that much value out of AI.” (Packy, 03:11)
• “It's a bull market in yarn spinning, folks. Get ready, get out the yarn and start spinning.” (Packy, 28:59)
• “Hackers used Claude to steal 150 gigabytes of Mexican government data.” (Chris, 25:23)
• “...Sell Claude subscriptions to aliens?” (Chris, 25:19)
• “It ain't easy having principles.” (Edward, quoted by Packy, 25:19)

Important Timestamps

• AI adoption study breakdown: 00:02–08:25
• AI & energy grid, Trump policy: 10:51–14:33
• Funding, venture capital trends, IPO scale: 16:43–20:28
• Anthropic safety policy shift & reaction: 20:28–25:49
• Perplexity Computer/AI product launches: 27:08–28:01
• Media/finance news flashes: 29:22–30:05

Tone & Style

The episode blends sharp skepticism with informed industry analysis, frequently poking fun at hype (and its viral incentives), questioning conventional wisdom, and dissecting the gray area between perception and data in tech. The hosts and guests bring a mix of irreverence, expertise, and a healthy mistrust of simple narratives—especially around AI, power politics, and market hype.

In Summary

This episode covers the thorny gaps between AI progress and corporate adoption, new regulatory and infrastructural challenges around energy use, and the eroding boundaries around AI safety among major labs. It’s required listening for anyone following the intersection of business, policy, and technology in 2026—with a strong dose of both data realism and tongue-in-cheek industry banter.