2025-04-04

Apple’s AI Crisis & ChatGPT’s World Domination w/ Dave Blundin & Salim Ismail | EP #162

Peter Diamandis hosts MIT-connected venture investor Dave Blundin (Link Ventures) and Salim Ismail for a rapid news-review episode covering OpenAI's $300 billion valuation debate and the X/xAI merger, the rise of ultra-young one- and two-person unicorn founders (Mercor's 21-year-old CEO), and Gemini 2.5 Pro's benchmark dominance alongside AI-pioneer Nobel prizes. They dig into Apple's Siri/AI crisis as a classic big-company innovator's dilemma, China and Estonia mandating AI in schools, and the shift from hand-coding to 'vibe coding.' The conversation also covers MCP as an emerging AI interoperability standard, Google's AGI-by-2030 safety paper and how to define AGI, flying taxis, the humanoid robot race (Figure, Atlas), and closes with Bitcoin's correlation to equities and CoreWeave's IPO pop as a possible reopening of AI capital markets.

▶ Watch on YouTube

Topics

OpenAI's $300B valuation and the X/xAI merger Economy ▶ 3:00

Dave Blundin and Salim Ismail debate whether OpenAI's valuation is justified, and discuss Elon Musk merging X and xAI to combine Twitter's data with AI.

Young AI-native founders and the one-person unicorn era Economy ▶ 11:33

Link Ventures now backs 30 companies a year (up from 10) as AI lets tiny, very young founding teams (like Mercor's 21-year-old CEO) reach multi-billion-dollar valuations in under two years.

ChatGPT's viral image-generation moment AI ▶ 22:57

A viral AI image feature drove a million ChatGPT signups in an hour; the panel discusses how AI-driven interactive content is pulling audiences away from passive media like movies and TV.

Apple's Siri crisis as a big-company innovator's dilemma AI ▶ 31:51

Apple's AI failures are framed as the classic problem of big-company control frameworks being too slow for the pace AI now demands, contrasted with Zuckerberg's ship-fast culture at early Facebook.

AI mandated in schools and the future of education AI ▶ 36:51

China and Estonia are baselining AI into primary/secondary education in fall 2025 while US schools still treat it as cheating; the panel discusses pull-based vs push-based learning.

Peter Diamandis's longevity stack Longevity ▶ 45:56

Diamandis describes his annual Fountain full-body diagnostic, Viome gut-microbiome testing, and OneSkin senolytic skincare as his three core health technologies.

Vibe coding and the end of traditional programming AI ▶ 48:28

Dave Blundin says Emad Mostaque's old prediction of 'no more coders' is arriving faster than expected; the current move is to build via AI vibe coding rather than hand-debugging generated code.

Gemini 2.5 Pro dominance and AI Nobel prizes AI ▶ 50:16

Gemini 2.5 Pro leapfrogs rivals and nears 20% on Humanity's Last Exam; the panel also discusses Demis Hassabis and Geoffrey Hinton's Nobel prizes and open-source vs closed models.

MCP as an open standard for AI-tool interoperability AI ▶ 59:25

Dave Blundin explains Anthropic's Model Context Protocol (MCP), now backed by OpenAI and Microsoft, as an HTML-like open standard letting AI safely access tools and data.

Defining AGI and Google's 2030 AGI safety paper AI ▶ 1:02:32

The panel discusses Google's AGI safety paper, Eric Schmidt's view that the real risk is a 'dystopian user' rather than US-China competition, and Demis Hassabis's 'artificial Einstein' AGI benchmark.

Flying taxis and eVTOL regulatory bottlenecks Transportation ▶ 1:06:53

China clears E-Hang flying taxis for commercial use while US eVTOL companies Archer and Joby remain stuck on regulation and infrastructure despite having the technology ready.

Humanoid robot race: Figure and Atlas Robotics ▶ 1:12:08

Figure's and Boston Dynamics' Atlas robots show rapid gains in human-like movement; the panel discusses why humanoid form factors are chosen and how learned skills transfer instantly across robot fleets.

Predictions made

open Dave Blundin: Within about two years, there will be multiple more examples of tiny teams reaching multi-billion-dollar valuations within two years of founding.

· due: 2027 · ▶ watch

“I suspect in two years we'll have multiples of that in terms of case studies to point to.”

Your call:

open Dave Blundin: A company run by just two or three people will reach a multi-billion-dollar valuation very soon.

· due: unspecified · ▶ watch

“there's going to be a company with two or three people that reaches a multi-billion dollar valuation sometime very soon”

Your call:

open Dave Blundin: Super-empathetic AI voice models like Sesame will be combined with top-tier LLM reasoning within one to two months.

· due: 2025-06 · ▶ watch

“that'll come out within the next I'm going to guess within the next month to two months”

Your call:

open Elon Musk: Short AI-generated video movies will become dominant very shortly, followed by full feature-length AI-generated films about a year later.

· due: 2026 · ▶ watch

“short video movies are going to be becoming dominant very shortly... full motion picture length next year”

Your call:

open Dave Blundin: Traditional hand-coding for most software will be 'game over,' replaced by vibe coding within 2025.

· due: 2025 · ▶ watch

“it's going to be game over within, you know, certainly within 2025 and then everything will move to vibe coding”

Your call:

open Dave Blundin: xAI's Grok will come out ahead long-term because Elon Musk controls chip manufacturing access, unique data from X and Tesla, and essentially unlimited capital.

· due: unspecified · ▶ watch

“I tend to bet on Grock. I just think he's got an unlimited checkbook”

Your call:

open Salim Ismail: An unknown, unforeseen entrant will have the highest score on the Humanity's Last Exam benchmark two years from this episode.

· due: 2027 · ▶ watch

“I think it's going to be an unknown entrant”

Your call:

open Dave Blundin: Archer Aviation and Joby Aviation plan to have eVTOL flying taxis operational in time for the Los Angeles Olympics.

· due: 2028 · ▶ watch

“the plan right now, as far as I know, is when the Olympics are coming in a few years here in LA to have them up and operational”

Your call:

open Dave Blundin: CoreWeave's strong IPO performance will help reopen both the IPO and M&A markets for AI companies broadly.

· due: 2025 · ▶ watch

“this is going to crack open not only the IPO markets, but also the M&A markets for all these AI companies”

Your call:

Numbers that matter

$20 trillion projected AI economic impact within a couple of years Dave Blundin on the scale of the AI economy despite valuation uncertainty
$40 billion raised by OpenAI, roughly half earmarked for compute Dave Blundin on OpenAI's latest funding round
Mira Murati's Thinking Machines founding valuation: about $3-4 billion Peter Diamandis and Dave Blundin comparing new foundation-model startup valuations
Ilya Sutskever's Safe Superintelligence founding valuation of about $9 billion, now reportedly $30 billion Same discussion of ex-OpenAI founders' new companies
Almost no public AI companies have more than $10 million in revenue Salim Ismail on why AI startups are staying private
Link Ventures backs about 30 companies a year now, up from about 10 historically Dave Blundin on incubator deal flow roughly tripling
Mercor, an AI hiring-interview startup founded by a 21-year-old, raised about $100 million Headline story discussed by Dave Blundin and Peter Diamandis
Mercor's valuation rose about $20 million per week after Link Ventures invested Dave Blundin explaining why he bought student housing near MIT
$5.5 million spent on an apartment building for founder teams near MIT Dave Blundin on Link Ventures' housing investment for startups
70-75% of incoming MIT freshmen want to start a company before they graduate Dave Blundin and Peter Diamandis on founder-age data trends
Portfolio company Blitzy writes about 3 million lines of code in a single night, roughly $300 million / 3 years of equivalent R&D Dave Blundin on AI-driven code generation speed
Gemini 2.5 Pro scores close to 20% on Humanity's Last Exam Dave Blundin on the hardest AI benchmark in existence
About 40% of oil and gas deposits in the last 20 years were found using digital/sensor search Salim Ismail comparing to the AI critical-minerals discovery story
Goldcorp's market cap rose from $50 million to $3 billion after crowdsourcing geological data for a $500,000 prize Dave Blundin telling the Goldcorp Challenge data-mining story
Carlsberg's open-sourced yeast led to a 10x increase in global beer consumption over the following 10 years Dave Blundin's open-source-history story

Worth digging into

🕳️ Demis Hassabis's AGI benchmark: an artificial Einstein deriving relativity from pre-1910 data

A concrete, codable AGI test that sidesteps endless definitional arguments about what counts as AGI.

🕳️ The Goldcorp Challenge (crowdsourced geological data mining)

A pre-AI proof that giving outsiders proprietary data finds value insiders missed through cross-domain pattern matching — a template now supercharged by AI, directly paralleling the TechCrunch critical-minerals story discussed earlier in the episode.

🕳️ MCP as a possible 'HTML moment' for AI tool access

Dave Blundin compares Anthropic's Model Context Protocol to early HTML and to the OAuth/Internet Identity Workshop lineage — worth tracking adoption by OpenAI and Microsoft as a possible internet-scale open standard.

🕳️ Humanity's Last Exam as the new frontier AI benchmark

Gemini 2.5 nearing 20% suggests this benchmark, not the Turing test, is now the real race to watch.

🕳️ Carlsberg's 1883 open-sourced yeast

A pre-digital, real-world case of open-sourcing a proprietary breakthrough leading to 10x industry growth — a useful historical parallel for the open-vs-closed AI model debate.