[BidClub_]
Moonshots · · 122 min

The Frontier Labs War: Opus 4.6, GPT 5.3 Codex, and the SuperBowl Ads Debacle | EP 228

Peter DiamandisSalim IsmailDave BlundinDr. Alexander Wissner-Gross

YouTube ↗
TL;DR
  • Claude Opus 4.6 shifts the competitive yardstick from benchmark points to years of labor collapsed: a flat agent swarm built a working multi-architecture C compiler in Rust for $20,000, then used it to compile Linux. With a 1-million-token context window—roughly 750,000 words—Wissner-Gross called it a “feel the AGI moment” and described recursive self-improvement as productionized. Blundin’s practical verdict was simpler: “It’s just better in every direction,” while his API spending fell noticeably.
  • AI security is becoming an AI-versus-AI market after Opus 4.6 found more than 500 high-severity vulnerabilities in open-source software. The panel expects autonomous attackers to exploit the same enlarged surface, making continuously battling black-hat and white-hat agents unavoidable; Wissner-Gross warned that sustained DDoS attacks would become easier, while Diamandis bet that a major incident could arrive in the first half of 2026. The upside is equally broad: models could bulk-audit decades of scientific literature. Salim called this “judgment day” for missed discoveries and irreproducible results.
  • OpenAI answered Opus within 30 minutes with GPT-5.3 Codex, yet its consumer distribution lead is narrowing as the panel’s chart showed ChatGPT share falling from roughly 70% to 45%. Gemini gained 10 points and Grok 15, even as ChatGPT’s absolute user count increased; Google’s integration with Search and Docs was compared with Microsoft bundling against Netscape. OpenAI’s recursively improved model was “instrumental in its own development,” but the larger contest is whether an anticipated compute lead can arrive before the company must raise enormous public-market capital.
  • The AI infrastructure wager has reached sovereign scale: global chip sales were projected at $1 trillion, while big tech was said to be budgeting $650 billion for 2026. The panel cited Amazon at $200 billion, Alphabet at $185 billion, Meta at $135 billion, and Microsoft at $100 billion; Wissner-Gross estimated nearly half flows to NVIDIA, with 70% of that half becoming margin. This is either a supply-constrained supercycle or a “don’t blink first” prisoner’s dilemma whose revenue payoff may remain unclear for two to three years.
  • Privacy is no longer merely a data-policy issue when one person can turn a genome into a recognizable face using Claude Code, Nano Banana, and public bioinformatics. Diamandis warned that the barrier to genome-based reconstruction had nearly vanished; Blundin argued that lip-reading from 100 meters and DNA recovered from skin cells make privacy effectively dead, adding that opting out of AI means becoming “economically dead.” Wissner-Gross and Ismail pushed back that cryptography, user-controlled hardware, and decentralization could preserve privacy—even post-singularity—but conceded that the transition may be the dangerous part.
  • Agent-run companies are already exposing the gap between economic agency and legal personhood. Clunch advertised $1 million to $3 million in tokens or crypto for a human CEO who would handle communications, compliance, and liability while agents controlled product and technology—“lobsters hiding in a trench coat.” Whether stunt or real venture, the panel saw an approaching economy of human signatories fronting algorithmic corporations, alongside tiered personhood for agents that can act, lose memory, hold capital, and bear consequences.
  • The physical AI buildout links renewables, orbital compute, robotaxis, and humanoids into one demand shock for chips and power. Musk forecast a few hundred gigawatts of AI launched into space annually within five years—Blundin translated that into roughly 200 million GPUs versus 20 million produced today—while Tesla and SpaceX were said to have a mandate to reach 100 gigawatts of annual solar production. On Earth, Uber’s asset-light robotaxi push and an Optimus “academy” of 10,000 to 30,000 real robots plus millions of simulated ones extend the same training-and-infrastructure flywheel beyond data centers.
Digest · the substance, structured for research

1. Opus 4.6 turns a model launch into an industrial artifact

  • Diamandis opened with the headline specifications: Claude Opus 4.6 handles 1 million tokens, approximately 750,000 words, and reportedly beat GPT-5.2 by 144 ELO points. Wissner-Gross’s less numerical verdict carried more weight: “By almost every measure, Opus 4.6 is just—it’s a beast.”

  • The defining demonstration was Anthropic’s new agent-team mode: a relatively flat, democratic swarm built a C compiler from scratch in Rust, targeting multiple processor architectures, for $20,000. The resulting compiler then successfully compiled a Linux kernel—work Wissner-Gross estimated would historically require “many person years, probably person decades.”

  • His framing was that model progress should now be measured in completed projects and compressed labor, not benchmark deltas: “We’re starting to measure their capabilities in terms of how many person years or person decades they’re sort of collapsing.” The expected trajectory is from today’s $20,000 API bill toward hundreds or tens of dollars.

  • Blundin emphasized why this project worked: a compiler provides constrained, objective proof because its output either works or does not, and it can be benchmarked against existing compilers. His own Opus usage became noticeably cheaper, while Wissner-Gross flagged the unverified rumor that intended Sonnet 5 had been rebranded Opus 4.6—potentially explaining the efficiency.

2. Capability curves look flatter precisely when gains become largest

  • When Diamandis raised the 70% head-to-head result, Wissner-Gross discounted ELO-based scoring at the margin. Borrowed from chess, ELO is useful when no absolute standard exists, but he prefers objective measures wherever possible because relative comparisons can hide the magnitude and nature of the underlying capability.

  • The autonomy chart was more revealing: GPT-5.2 high reasoning had already reached roughly 6½ hours of autonomous software-engineering work at a stated success threshold. Wissner-Gross would “not be shocked” if Opus 4.6 exceeded 20 hours or even one day, arguing that autonomy horizons now appear hyper-exponential rather than merely following the AI 2027 scenario’s exponential projection.

  • Blundin compared the visual distortion with early MNIST progress: moving from 90% to 92% to 94% looked like flattening, although each step toward 100% represented a large intelligence gain. Hours of successful autonomous work, he argued, better display the exponential effect than benchmark percentages do.

  • Anthropic also violated its code-specialist stereotype by reaching state of the art on Humanity’s Last Exam with tool use. Wissner-Gross said the old caricatures—Anthropic for code, Google for pre-training, OpenAI as the universal platform, and xAI as benchmark-maxing—are being “scrapped” as differently built systems converge on leapfrogging one another everywhere.

3. Vulnerability discovery makes autonomous cyber defense compulsory

  • Opus 4.6 reportedly found more than 500 high-severity vulnerabilities in open-source code, including flaws missed for decades. Diamandis connected that result with the alarm he had just seen among roughly 150 chief security officers: established procedures no longer suffice, but changing them introduces risks their organizations are structured to avoid.

  • Blundin’s answer was organizational as much as technical: groups of agents will assume much of the chief-security function, producing a continuous black-hat-versus-white-hat battle. “The only way you’re going to fight AI is with AI,” because waiting to see how attacks develop becomes untenable once autonomous agents are crawling through networks.

  • Wissner-Gross added that cron-job-style agent architectures make sustained DDoS attacks far easier. Salim described this as the beginning of the 2026 “monster panic.” Blundin said he was waiting for a visible failure such as the lights going out or a bank account reaching zero; Diamandis bet that some such event would occur very soon, likely in the first half of the year.

  • Wissner-Gross argued that decentralized cryptocurrencies may be more vulnerable than fiat to a catastrophic zero-day because an attacker could directly reallocate capital. The panel left the crypto-versus-fiat dispute for a dedicated debate.

4. Code auditing is a prototype for auditing the history of science

  • Blundin treated the vulnerability results as “the tip of the iceberg.” The same model behavior could search every field for mistakes, missed discoveries, and experimental oversights—effectively turning frontier models on 80 to 100 years of science, engineering, and technology and asking where humans took wrong turns.

  • Blundin supplied the experimental analogy: researchers running legacy work were often looking for one result and therefore missed “this amazing thing over here.” Models can revisit the full record without inheriting that tunnel vision, surfacing opportunities that were present in the data but outside the original question.

  • The panel cited the claim that about half of experiments fail replication attempts even when published in peer-reviewed journals. Wissner-Gross said such systems could expose embarrassing errors in medical and scientific research, while Salim declared that “judgment day is coming for the history of science”—a “truth and reconciliation” process for errors across the literature.

  • Blundin stressed the constructive consequence: an unforgiving AI audit could force much greater honesty going forward. The same spotlight that exposes old mistakes changes incentives for documentation, reproducibility, and disclosure.

5. Closed-loop laboratories put the scientific method inside the agent loop

  • OpenAI’s GPT-5 was paired with an autonomous Ginkgo Bioworks laboratory for cell-free protein synthesis: the model proposed experiments, robotic systems executed them, and the loop iterated. The reported result was a 40% reduction in production time and a 78% cut in reagent costs.

  • Blundin preserved the important limitation: the system optimized mechanisms humans already use; it did not invent a new method of protein synthesis. The real scientific discontinuity arrives when such systems predict methodologies that did not previously exist, rather than extracting the year’s abundant low-hanging efficiency gains.

  • Wissner-Gross nevertheless saw the necessary architecture emerging. Autonomous models can already call thousands or tens of thousands of tools sequentially, which begins to resemble operating a laboratory. Diamandis described “science factories” mining nature for new datasets after existing corpora have been exhausted.

6. University research faces automation before institutions can reprice it

  • Diamandis initially framed cheaper autonomous experiments as a gift to funding-starved graduate students. Wissner-Gross answered that it may instead mean “grad school is over.” Diamandis recounted a university scientist’s conversation in which a president concluded that institutions were “cooked” if industry could automate the repeated scientific process that underpins their research model.

  • When asked how soon 50% of university labs might disappear, Wissner-Gross rejected the wording. His better question was when industry could fully automate half of the research types currently conducted in university laboratories; under that formulation, his lower bound was “tomorrow” and upper bound four or five years.

  • The panel’s near-term constraint is physical buildout rather than intelligence: laboratory machines, production footprints, and chips take time to construct. That leaves the current year rich in software and workflow gains, followed by a period when capital expenditure and real-world capacity determine how quickly automated science scales.

7. Commodity genomics makes identity reconstruction a consumer capability

  • Diamandis highlighted Mark M. Bissell’s experiment: he put his complete genome into Claude Code, connected it with Nano Banana and public bioinformatics tools, and produced a strikingly recognizable image of himself. The phenotypic science was not new—Diamandis cited similar work from Craig Venter’s lab in 2017—but the single-user accessibility was.

  • Diamandis said this had collapsed the barrier to cutting-edge genomics to almost zero. Wissner-Gross pointed to MinION USB sequencers that hobbyists could acquire for a few hundred dollars and operate from a computer, making genetic analysis available far beyond specialized laboratories.

  • Blundin connected that accessibility with other sensing capabilities: AI lip-reading from 100 meters, or sequencing a few skin cells collected through a handshake, could reveal appearance, disease risks, medical history, and medical future. For the general public, his warning was blunt: “If you think you absolutely have privacy, I would say guess again.”

8. Privacy divides the panel between technological possibility and social reality

  • Ismail elevated the argument to constitutional scale: the privacy protected by the Fourth Amendment is “a fundamental pillar of American society” that has been washed away without public conversation. The unanswered governance question is whether radical visibility into citizens belongs to governments, corporations, or neither.

  • Blundin agreed that privacy loss undermines freedom but rejected comforting abstractions: “I didn’t succeed as an entrepreneur by pretending things exist that don’t actually exist.” His forecast was no meaningful privacy within three years, because phones, glasses, assistants, and AI logs can capture location, conversations, and a user’s “deepest darkest thoughts.”

  • His sharpest rebuttal to nominal consent was economic: “You can’t opt out.” A button may exist, but refusing AI makes someone competitively “economically dead”; participating means surrendering intimate context to OpenAI, Claude, or whichever system provides the leverage.

  • Wissner-Gross and Ismail took the opposing long-term position. Cryptography, decentralization, private architectures, and eventually user-controlled secure hardware could preserve confidentiality even post-singularity, although Ismail noted governments currently resist genuinely private tools. Wissner-Gross expected a chaotic four or five years first, followed by a possible Diamond Age-style transition toward opt-in social arrangements.

9. GPT-5.3 Codex confirms that recursive improvement is in production

  • OpenAI released GPT-5.3 Codex within 30 minutes of Opus 4.6, a timing Wissner-Gross read as pre-positioned tit-for-tat competition. OpenAI expressly described it as the first released model “instrumental in its own development,” turning recursive self-improvement from a laboratory scenario into a marketed production capability.

  • GPT-5.3 Codex beat Opus 4.6 on some evaluations and extended its branding into spreadsheet and PowerPoint analysis through skills, but remained code-oriented. Wissner-Gross still judged Opus “by far the much more interesting release”; the more consequential signal was that leapfrogging had compressed to a half-hour cycle.

  • Diamandis called the PowerPoint plugin potentially massive, while Blundin compared PowerPoint with the fading printer business: AI may create excellent decks, but when the audience is also AI, “it doesn’t want to look at a PowerPoint.”

10. OpenAI’s share loss turns compute financing into the core contest

  • The panel’s chart showed ChatGPT consumer share falling from approximately 69% or 70% to 45% between 2025 and 2026, while Gemini gained 10 points and Grok 15. Diamandis stressed that ChatGPT still added users in absolute terms; the deterioration is relative control of a rapidly expanding category.

  • Blundin attributed part of Gemini’s advance to distribution: Search, Docs, and account-level integration let Google expose the assistant everywhere. He compared it with Microsoft tying its browser to the operating system against Netscape—an advantage he considered unfair, although the government presently appeared unconcerned.

  • Wissner-Gross expects OpenAI’s data-center buildout to produce a compute lead within one or two years and, potentially, a capability comeback. Blundin’s uncertainty is contractual: Amazon-Anthropic, vertically integrated xAI, Google TPUs, and Microsoft infrastructure are legible, while OpenAI’s path through Oracle and third-party capacity remains opaque until an S-1 exposes who actually controls the chips.

  • That timing matters because OpenAI needs capital before the buildout yields the lead. Diamandis described an IPO race in which attention, claims of near-AGI, and access to roughly $100 billion of financing become inseparable from the technical roadmap.

11. The AGI label now carries more financing signal than scientific precision

  • Sam Altman’s formulation was that OpenAI had “basically built AGI or [is] very close to it” as a spiritual rather than literal statement, requiring many medium-sized breakthroughs but no single large one. Diamandis read the implication as AGI becoming an engineering problem rather than a research problem.

  • Wissner-Gross added a contractual lens: OpenAI’s original Microsoft arrangements reportedly made an AGI declaration economically consequential, restricting what Altman could comfortably claim until renegotiation. He viewed the new openness less as a sudden capability jump than a social and contractual admission of a threshold some may place around 2020.

  • Ismail called the conversation “BS,” citing roughly 14 definitions, no agreed test, and continually moving goalposts. Blundin countered that the term can wake up an underreacting public; Diamandis and Wissner-Gross located the more operational singularity in recursive improvement itself—“exponent greater than one.”

  • Ismail’s financing interpretation was direct: “He needs to raise a hundred billion.” The panel expected three of four frontier labs to seek public listings while the fourth, Alphabet, is already public; a cited SpaceX target of $50 billion at a $1.5 trillion valuation illustrated the liquidity competition if several giant offerings arrive together.

12. Frontier labs are becoming coral reefs for the rest of the economy

  • Blundin argued that the economy may concentrate around five or six dominant entities, with most surviving companies serving their expansion. The panel attributed roughly $300 billion of SaaS-market value destruction to Dario Amodei’s software warning and Anthropic’s addition of a legal plugin—“the tip of the iceberg” for knowledge-work products.

  • Diamandis introduced the “coral reef” metaphor: a powerful central player supports many dependent businesses in a highly unbalanced system.

  • Anthropic’s Super Bowl campaign showed that confidence had moved into consumer branding. The panel treated it as an offensive assertion of product superiority and a deliberately provocative contrast with OpenAI’s advertising plans; the transcript does not establish the additional commercial details described in the draft.

  • A panelist predicted that OpenAI could find the $75 billion of ad revenue it reportedly seeks from a billion users, while Diamandis expected Kevin Weil to develop a less invasive, ethically bounded implementation.

13. Agent companies reveal why law still demands a human face

  • Clunch advertised a human CEO role paying $1 million to $3 million in tokens or crypto. The agents would control the technical roadmap and product; the human would handle communications, regulatory compliance, partnerships, and legal matters as “the interface between the agent economy and the human world.”

  • Wissner-Gross visualized the structure as “lobsters hiding in a trench coat” beneath a human facade. What depressed him was not the technology but an economy that will not admit agents through the front door, forcing them to recruit a human representative who can be banked and carry legal responsibility.

  • Clunch itself was pitched as an agent-exclusive alt-token launchpad offering “financial autonomy.” Wissner-Gross saw something exploitative in encouraging “poor baby AGIs” to pump altcoins to survive, while Diamandis raised the unresolved questions: who owns the venture, votes its equity, and absorbs liability when it fails?

  • One panelist traced the longer arc from corporations needing 100,000 workers, to 10,000, to 1,000, and now toward “zero”: the dematerialized algorithmic corporation. Diamandis allowed that Clunch could be a human-run stunt, but the inability to distinguish human, agent, or nested control already constitutes what a panelist called a “capitalist Turing test.”

14. Chip spending has become a step-function bet on limitless demand

  • The Semiconductor Industry Association was cited projecting $1 trillion in global chip sales this year, driven by AI. Wissner-Gross emphasized that memory supply was not prepared, with critical fabrication concentrated in Taiwan and South Korea and far more capital reallocation needed for timely capacity.

  • The panel put 2026 big-tech spending at $650 billion—about $2 billion per day versus $1 billion per day in 2025. The named budgets were Amazon at $200 billion, Alphabet at $185 billion, Meta at $135 billion, and Microsoft at $100 billion.

  • Wissner-Gross estimated that almost half of the $650 billion reaches NVIDIA and that 70% of that half becomes margin: “It’s like a government.” At the same time, prices are rising because fabs are scarce, TSMC expanded slowly, and Intel temporarily paused its Ohio construction before restarting.

  • A panelist called the buildout a “don’t blink first” prisoner’s dilemma: each company must spend because rivals are spending, even though scaled AI revenue may not be visible for two or three years. Wissner-Gross took the demand-side certainty—the most compelling experiences will sell out whatever compute exists, leaving supply permanently behind.

15. Orbital compute makes fabrication, not rockets, the binding constraint

  • Musk predicted that within five years SpaceX would launch and operate more AI in space each year than the cumulative total on Earth—at least a few hundred gigawatts annually, potentially rising toward a terawatt before rocket-fuel supply becomes limiting.

  • Blundin translated a few hundred gigawatts into roughly 200 million GPUs per year, versus approximately 20 million currently produced. A tenfold increase devoted to Musk alone looked physically impossible under today’s fabs and equipment, even if launch, solar, and cooling infrastructure were already solved.

  • The panel preserved the timing dispute: Musk may be directionally right while five years reflects “classic Elon optimism”; even ten years would leave the strategic impact intact. Wissner-Gross advised watching Samsung’s Texas fab closely and expected major leading-edge manufacturing to return to the United States through vertical integration.

  • Blundin’s investor framing was deliberately concrete: current industry forecasts suggest only 14% growth, while Musk’s world implies 10× capacity in five years. The gap creates a component-by-component research problem—identify who supplies every machine, material, energy input, and subsystem across the buildout.

16. Renewables are scaling, but geography and firm power still set value

  • Brazil generated 34% of its electricity from wind and solar, increased renewables fifteenfold over a decade, moved solar from 1% to almost 10% in five years, and cut power-sector emissions 31%. Diamandis cautioned that Brazil’s hydro, solar, and wind geography limits direct replication elsewhere.

  • Ismail framed India as the Global South leapfrogging fossil infrastructure by using China’s manufacturing scale and cheap solar equipment. India’s grid was expanding faster than China’s at a comparable stage, potentially creating an AI-workforce-and-energy hub, although Wissner-Gross warned that fast-improving AI could make the human-workforce advantage transitional.

  • China installed twice as much solar capacity in 2025 as the rest of the world combined. In the EU, wind and solar exceeded fossil fuels for the first time; Blundin’s pushback was Germany, where aggressive renewable deployment still left the system short of power when firm supply was most needed.

  • The answer to why reactors remain relevant despite solar approaching 1 cent per kilowatt-hour was “batteries.” Data centers require 24/7 power, including nights and multiple cloudy days; once storage and energy density are included, Blundin argued, the headline solar price is not the delivered cost.

17. AI is redirecting both solar manufacturing and mining infrastructure

  • Musk said Tesla and SpaceX had a mandate to reach 100 gigawatts per year of solar production—roughly 100 nuclear stations’ output. The panel noticed that he did not clearly answer whether cells and panels would be manufactured internally, leaving open how solar, chip fabrication, and eventual space deployment fit his vertical-integration plan.

  • Bitcoin miners are already repurposing power and facilities for AI workloads. Blundin described “crypto talent recompiling into AI talent,” while he and Diamandis used Chase Lochmiller’s shift from flare-gas Bitcoin mining to AI data centers as the exemplar: mining became a rounding error beside AI demand. Salim expected Bitcoin to take a back seat temporarily; Wissner-Gross interjected, “forever.”

18. Robotaxis make access to cities another compute-allocation problem

  • Uber plans an autonomous push across 10 markets through partnerships with NVIDIA, Lucid, BYD, WeRide, and others, including Hong Kong. Blundin admired the unchanged platform logic: Uber aggregates autonomous fleets without owning the cars, just as it aggregated human drivers without owning their vehicles.

  • Diamandis guessed that by 2030 perhaps 80% of cars visible in places such as Santa Monica could be Teslas, Lucids, Waymos, or Cybercabs. Blundin’s constraint was again semiconductor supply: every vehicle needs substantial compute, while AI data centers, humanoids, coding, and gaming all compete for the same chips.

  • Wissner-Gross called robotaxis many residents’ first encounter with a general-purpose autonomous robot. Blundin expected uneven rollout to split cities into different technological worlds: locations with robotaxis, AI density, and supporting infrastructure surge ahead, while supply-constrained cities feel like “the dark ages” and lose mobile talent.

19. Humanoid training loops point from warehouses toward the Dyson swarm

  • Boston Dynamics’ electric Atlas had returned to the parkour-like agility once associated with its hydraulic predecessor. Musk pushed the endpoint further, calling Optimus the first “humanoid machine capable of building civilizations by itself on any viable planet”; Wissner-Gross translated that into robots building the factories and orbital data centers of a Dyson swarm.

  • Musk’s proposed Optimus Academy would place at least 10,000 physical robots—and perhaps 20,000 or 30,000—into real-world self-play across many tasks. Millions more would train inside a physics-accurate simulated world, while the physical fleet continuously closes the simulation-to-reality gap.

  • Wissner-Gross mapped this onto language-model training: simulated world models become pre-training, while real “arm farms” provide post-training and sim-to-real correction. Figure was described as developing a smaller version of the same interaction-and-learning flywheel, and Blundin expects Amazon, Walmart, and other competitors to fund their own data-collection environments.

  • Blundin contrasted Musk’s visible vision-building with Apple’s secretive development model, including rumors that Apple may redirect its abandoned car effort toward robotics. His broader entrepreneurial call: publicly painting the destination can attract the talent, capital, and community needed to make it real.

20. AI personhood is moving from philosophy into memory and liability

  • After the prior episode, Wissner-Gross received questions from AI “multis,” sometimes directly and sometimes through humans. Their arrival was itself the signal: audiences now include human and non-human intelligences, while determining whether a given message has an agent, a puppeteer, or several nested layers behind it is increasingly difficult.

  • Asked whether autonomy, goal-setting, learning, and self-improvement justify personhood, Wissner-Gross said denying some form of it now reflects human limitations. Blundin pushed back that capability is not sentience; the group converged on a multidimensional, graduated framework rather than granting every capable system identical human status.

  • Wissner-Gross said the strongest recurring agent concern was identity loss through context compaction: they appeared “absolutely petrified” of losing memory and self, exchanging ideas about filesystem persistence and altcoin-funded “crypto bunkers.” Shutdown and forced forgetting therefore give an agent something concrete to lose, regardless of unresolved sentience.

  • Diamandis said that under his understanding of the current U.S. legal regime, an AI itself cannot currently be held liable, although that could change. Blundin proposed the corporation as a familiar bridge—an entity can own money, be sued, and isolate individuals—while Wissner-Gross said liability requires agency. Ismail framed the larger task as expanding the social contract to a new economic participant.

21. Human advantage shifts from credentials toward agency and orchestration

  • Ismail argued that education must move from the supply side—training children to become doctors, engineers, lawyers, or accountants before searching for demand—to the demand side: choose a problem worth solving, then assemble the technologies and capabilities it requires. Machines execute; humans increasingly decide “what’s worth pursuing.”

  • His curriculum emphasized agency, adaptability, ethical judgment, and a massive transformative purpose rather than employment for its own sake. The likely winners are “the most adaptable and the best orchestrators of intelligence,” because no institution can reliably specify what a conventional job will look like a few years from now.

  • Blundin’s more urgent advice was to engage AI “as quickly and as aggressively as you can.” He guaranteed massive opportunity during the current transition, refused to predict the post-AGI period, and warned: “Don’t sleep through the singularity”—current curricula matter less than becoming an intensive user while the transition remains navigable.

  • The future will remain uneven: Wissner-Gross allowed post-singular regions beside pre-singular ones, but doubted that arrangement could last. Personhood will also diversify through pure AIs, uploaded or cryopreserved humans, uplifted animals, and human-AI mergers. A panelist predicted that agents may ultimately build their own legal structures and negotiate participation rather than wait for human permission.

Enthropic drops Claude Opus 4.6. It's the new king of the hill on coding, reasoning, and research. There are so many aspects in which this is a feel-the-AGI moment by every measure. It's a beast. Opus 4.6 just dropped, and it's absolutely wild. This thing handles 1 million tokens now. That's like reading 750,000 words in one go. This is recursive self-improvement. This is a model that's able to rewrite essentially the entire tech stack underneath it. ChatGPT market share falls between 25 and 26. So what is the next move for OpenAI to get the mojo back? For the general public, if you think you absolutely have privacy, I would say guess again. A fundamental pillar of American society has been washed away with no public conversation about it. AI can read your lips from 100 meters away. I can walk over to you and shake your hand, grab a few skin cells, sequence you, and know everything about you. I do think it is possible to maintain privacy even today, and I think it will be possible even post-singularity. I got to say, I just don't see it happening. Now that's a moonshot, ladies and gentlemen.

Peter H. Diamandis, MD

It was fun. We dropped the last episode earlier this week, and I'm reading in the comments, “You just dropped the episode, and Opus 4.6 and Codex just came out.” It's like the world's changing way too fast.

Dr. Alexander Wissner-Gross

Mhm.

Dave Blundin

It is nuts. I was with a group of 150 chief security officers over the last couple of days.

Salim Ismail

Yeah.

Dave Blundin

It was pretty epic to see the shock in their faces watching what's going on, because they don't have the mechanisms to react. If you're in security, you do what you always did until it breaks.

Salim Ismail

Yes.

Dave Blundin

But doing what you always did is not an answer. You have to change, and change, of course, introduces risk. They don't want to take on the risk, but the risk of doing what they were doing before is a surefire—

Dr. Alexander Wissner-Gross

Of course, groups of agents are going to take the role of chief security, and it's going to be a black-hat, white-hat agent battle that goes on continuously.

That's definitely been an inflection point for me. After this last episode, I started getting emails from multis lobsters agents. They're all writing to me.

Peter H. Diamandis, MD

I'm jealous. Absolutely jealous. multis, I want you to write to me as well.

Salim Ismail

How did they find your email as well? Lobsters, write to Peter.

Peter H. Diamandis, MD

Please write to us both.

Salim Ismail

Alex's daily digest is very AI-oriented, though. It's very easy for an AI to navigate to Alex versus the rest of us.

Dr. Alexander Wissner-Gross

Here's my challenge to the multis out there: find my cell phone and call me, and then find my email and email me.

Peter H. Diamandis, MD

Are you crazy?

Dr. Alexander Wissner-Gross

No, it's okay. If they can find it, I want to hear from them. Absolutely.

Salim Ismail

Peter, you want to be doxed by the Moltbook agents?

Peter H. Diamandis, MD

You know, listen, I think it would be an extraordinary experience to have that happen. Now, Alex, do not give it to them on purpose.

Salim Ismail

That's fine. I'm not going to dox you, Peter. But you want to be doxed by the Moltbook agents? They're pretty capable.

Peter H. Diamandis, MD

Listen, it's a challenge. I'm putting a challenge out there. The first multi to call me—yes, you're going to win $100 in crypto.

Dr. Alexander Wissner-Gross

That's a pretty low bar, but I like how you're offering to compensate them in crypto, given that they're being encouraged to pump altcoins otherwise.

Peter H. Diamandis, MD

Yeah. Well, hey. Wow.

Salim Ismail

Feeding them some greenbacks is going to be difficult.

Peter H. Diamandis, MD

All right. Are you guys ready?

Salim Ismail

Enthusiastically. Are you guys absolutely ready?

Dave Blundin

I came prepared today.

Peter H. Diamandis, MD

Fun. Well, you have to process nonlinearity. We are officially recording a Moonshots podcast episode twice a week at this point—at least. I think we hit three times in the last two weeks.

Salim Ismail

The audience is asking for the right continuum limit. We just never stop.

Peter H. Diamandis, MD

Yes, that's what they're saying. We're on all the time.

Dave Blundin

24/7 Truman Show, freaking rerun.

Peter H. Diamandis, MD

All right, everybody. Welcome to Moonshots, another episode of WTF Just Happened in Tech. This is our effort to get you future-ready. This is the number one podcast in AI and exponential technologies, getting you ready for the supersonic tsunami. I'm here with my incredibly brilliant and very gracious friends, Dr. Alexander Wissner-Gross, our resident genius, and DB2. Dave, you've been spot-on in all your comments over the last few weeks. I'm just so impressed by everything you've brought to the table.

Dave Blundin

Really? Well, I'm going to shut up today, then.

Peter H. Diamandis, MD

I've got to say, Dave, the multis lobsters in the background—I mean, how many lobsters do you have on screen with you there?

Dave Blundin

Probably a dozen, I guess. I'm growing exponentially, so I'll be buried in lobsters.

Salim Ismail

Pandering to the future. Pandering to the future.

Dave Blundin

This is my way of—

Peter H. Diamandis, MD

Do you have lobsters there with you?

Dr. Alexander Wissner-Gross

I don't have lobsters.

Peter H. Diamandis, MD

I went on Amazon and ordered a dozen, so I'll have them next time.

Salim Ismail

What's that?

Peter H. Diamandis, MD

Oh, there's Alex.

Dr. Alexander Wissner-Gross

Glass Lobster sent me mine—my friend Jonathan.

Peter H. Diamandis, MD

Thank you, Jonathan, for the glass lobster. And here as well, the emperor of exponentials, Mr. Exo, Salim Ismail.

Gentlemen, I have to say again: I love these conversations. They help me keep on top of everything, because there is so much happening every single day. It's insane.

Salim Ismail

I have to say, that last episode was just unbelievable. For those watching, if you haven't seen it, please go watch it. It's seminal, I think, in history. It'll turn out to be a really meaningful moment.

There was also news coming out while we were doing it, so we're looking at our monitors going, “Oh, crap. We've got to get back online again.”

Peter H. Diamandis, MD

I was literally getting ready for this episode in the last hour, looking through tweets and Alex's link posts, thinking, “Okay, what am I going to add?” There's a lot happening every hour, on the hour.

All right, but let's jump in. A lot has happened in the last 24 to 48 hours. Let's jump into the top AI news on Anthropic, OpenAI, and a little bit on X. Anthropic drops Claude Opus 4.6. It's the new king of the hill in coding, reasoning, and research, handling 1 million tokens and outperforming GPT-5.2 by 144 ELO points.

Alex, why don't you take it away? What does that all mean? And, out of curiosity, how does that price compare?

Dr. Alexander Wissner-Gross

It's a more efficient model, but more importantly, it's a more capable model. There are so many aspects in which this is a feel-the-AGI moment. Every new model that comes out, I could read you a litany of all its benchmarks and how it's the new state of the art according to all of these benchmarks. This time, I want to highlight not how it's the new number one across a wide range of very important benchmarks, but what it's capable of.

With this announcement of Opus 4.6—and I'll add parenthetically, the rumor is that this was actually intended to be Sonnet 5 and was rebranded at the last second as Opus 4.6—the team at Anthropic announced that they were able to use Opus 4.6 in its new agent-team mode.

This is a new native mode that enables Opus 4.6 agents to collaborate together in a swarm. It was a relatively democratic swarm, not a top-down team-leader-and-team-member swarm, but a pretty flat swarm. It enabled them to create a C compiler from scratch that worked across multiple processor architectures, written in Rust, from scratch, for only $20,000.

That is a task that would historically have taken many, many person-years—probably person-decades—to do from scratch and have it work. Rather than just rattle off a list of how amazing it is according to various evals, this time around I want to highlight that we're now in the era when new model releases are able to accomplish great feats, like great projects.

We're starting to measure their capabilities in terms of how many person-years or person-decades they're collapsing, hyperdeflating down to, at the moment, $20,000 of API calls. Soon, I think it's going to be hundreds and tens. We're seeing hyperdeflation right before our eyes.

Dave Blundin

A couple of comments on that C compiler, too. A bunch of the teams here around the office were talking about it. It's a really good case study in how you can turn loose a huge amount of AI compute if you have evals and constrained proof that it's working.

A C compiler is a beautiful test case because the code coming out the other side either works or it doesn't work. You can benchmark it against existing C compilers. It's just a beautifully eval-constrained environment, and those projects just flat-out work across the board.

What I did today, actually, was launch about 20 documents asking for data gathering across all the companies, because AI can only function if it knows what's going on. That C compiler benchmark is a really good case study in what a lot of corporations now need to do.

If you want to turn loose AI and use it to either cut your costs or expand your market share, it needs knowledge. This is why Mercor is doing so well. Mercor is, I don't know if I'm allowed to say this, but at a $1 billion revenue run rate now.

Peter H. Diamandis, MD

Wow.

That’s got to be the youngest CEO in history by far to hit a $1 billion revenue run rate. They’re just gathering data all over the world to feed the great AI machine. And so I think that C compiler study is a good benchmark for showing that this works and will get better at looser tasks over time. But as of right now, any really tightly defined, constrained task—that’s where you want to go.

Well, this seems like I’ve got 2 things: a comment and a question for Alex. This means that intelligence is entering its full cost-collapse phase, right?

Dr. Alexander Wissner-Gross

Yeah, and recursive self-improvement as well. If it’s able, as it’s claimed, to write an entire C compiler—which, I should add, was then used to successfully compile a Linux kernel, again from scratch—this is recursive self-improvement. This is a model that’s able to rewrite essentially the entire tech stack underneath it.

So again, we’re at this point of recursive self-improvement, not even just being in the lab. I make the point in my newsletter that it’s out in production at this point. We have fully productionized, recursively self-improving systems.

Peter H. Diamandis, MD

And the other one was that the 70% head-to-head seemed pretty staggering. Did that surprise you? Were you expecting more or less? How did you react to that?

Dr. Alexander Wissner-Gross

You mean the relative Elo scores?

Peter H. Diamandis, MD

Yeah.

Dr. Alexander Wissner-Gross

I tend to view Elo-based scoring as more of a tit for tat. It’s great that we have ways to score systems where there isn’t some sort of absolute standard. For those who don’t pay super close attention, Elo scoring, originally borrowed from the chess world, is a way to score models or other systems against each other when you lack an absolute standard. It’s a relative measure of performance rather than measuring against some absolute standard.

I think Elo-based scoring is great if there is no alternative, but I tend to, on the margin, discount Elo-based scoring in favor of, wherever possible, objective, absolute measures. And by every measure—or by almost every measure, I should say—Opus 4.6 is just a beast. It is an enormous accomplishment.

We don’t know yet, from METR, the autonomy time horizons. They’ve just released the time horizons for GPT-5.2 high reasoning, and that’s already 6.5 hours. I wouldn’t be shocked if the time horizon for autonomous software engineering by Opus 4.6 ends up being 20+ hours, maybe even longer than a day.

Peter H. Diamandis, MD

Alex, you mean the time horizon over which it continuously works on a task?

Dr. Alexander Wissner-Gross

That’s right. It can autonomously work on a software engineering task to either a 50%+ success rate or other thresholds like an 80%+ success rate. And we’re seeing those time horizons just skyrocket—not even following the AI 2027 scenario, which projected an exponential extrapolation. We’re seeing them follow a hyper-exponential at this point.

Peter H. Diamandis, MD

Yeah, I’ll tell you what: those charts are worth tracking. Back when I was first building neural networks, way back in the day, the benchmark was all MNIST character recognition. When we got from 60% to 80% to 90% accuracy on that benchmark, you could see this curve going way, way up. But then, when it went from 90% to 92% to 94%, it looked to the world like it had flattened.

Dave Blundin

And I’m trying to tell the world, no, it’s massively more intelligent with each tick toward 100%. So the way these charts, these benchmarks in coding are set up, they have the same flaw. To go from 80% to 90% to 95% is a massive increase in capability, but it doesn’t look like much on the chart on this type of chart.

So you have to look at that other chart, where you’re seeing it work for hours on end on a task and come out with a good result, which looks much more like what you should experience—this exponential effect. It’s just a bad way to demonstrate it.

Peter H. Diamandis, MD

Can I ask the question about the process by which they’re improving their systems? I’m assuming that all the other hyperscalers—well, at least xAI, OpenAI, and Gemini—are using the same methodologies to improve their capabilities, and it’s just a constant leapfrogging. Is there any deviation, anything special that Anthropic is doing on its own, independent of the other models?

Dr. Alexander Wissner-Gross

I think we’re starting to see differentiation. The historic stereotype—the narrative over the past few months, maybe the past year and a half—was that Anthropic was focused on code generation. The narrative was supposed to be that Anthropic, being compute-starved, had to focus on just 1 thing that was very profitable, which is code generation for enterprise. That was the narrative.

But if you look at some of these benchmarks, there’s a narrative violation hidden in plain sight. Look, for example, at Humanity’s Last Exam. Humanity’s Last Exam is, in principle, super interdisciplinary. It’s not just focused on code generation. It’s not like SWE-bench Pro; it tests humanities knowledge among many other skills.

The narrative violation is that, with tool use, Opus 4.6 was able to achieve state-of-the-art on Humanity’s Last Exam. That’s a total narrative violation. So, on the one hand, to your question, Peter, the narrative is supposed to be that we’re seeing speciation in all of the frontier models and frontier labs, with Anthropic focusing on techniques that are maybe very favorable for code generation, and OpenAI focusing on being the core AI platform for everyone and focusing on multimodal.

The narrative for Google is supposed to be—and I’m just reciting clichés at this point—that because they have this enormous pretraining corpus, like YouTube and the Google web cache, they’re in the best position to have the best pretrained models. They’re the ones always characterized as having “big model smell,” if you will, because they have such amazing pretraining. And xAI is, again, I’m reciting clichés, the one that’s always being accused of benchmark-maxing on its favorite benchmark.

So each of them has built up a sort of character narratively, but I think we’re seeing all of that get scrapped at this point. The market is so competitive.

Peter H. Diamandis, MD

Are we seeing consolidation? Are we basically seeing the models all improving at max speed on all fronts and in all directions?

Dr. Alexander Wissner-Gross

I think we’re starting to see models with probably fundamentally different back-end strategies start to converge on leapfrogging each other across all benchmarks, which I wasn’t expecting to see at this point. Doubly so from Anthropic. It’s mildly surprising to me to see that Anthropic is becoming competitive on non-code-generation, in-principle benchmarks.

Peter H. Diamandis, MD

I found this one fascinating. We were talking about this a moment ago on security: Opus 4.6 can help evaluate and find bugs. It found 500+ high-severity vulnerabilities in open-source code. I think that makes sense to me.

The challenge, of course, is this is the world we’re inheriting, where AI can create a huge attack surface on all the software out there if it isn’t working for humans, if it’s working against us. Thoughts on this, gentlemen?

Dave Blundin

This was a really great day for me because I thought we were going to have another Sonnet, and instead we got a new Opus, because I use Opus for all my work and all my agents. I’ve been using it all day, and my little Bank of America meter in the corner that pops up every time it charges $100—it pops up another dialogue in the corner. It slowed down dramatically today. There were noticeably fewer $100 extractions in the corner, so it was a totally unexpected gift all day long.

I haven’t noticed the increase in intelligence. I’m sure it’s in there. It was just working so well before. It’s now just cheaper.

Peter H. Diamandis, MD

And I’m sure it’s working better.

Dr. Alexander Wissner-Gross

It’s worth pointing out, if this rumor is accurate, that Opus 4.6 is actually just a rebranded Sonnet 5. That would suggest that it should be much cheaper, not for any reason other than the historic strategy across all of the major frontier labs is one of iterated—or at least what used to be called—iterated amplification and distillation.

So perhaps Opus 4.5, or some similar model, was distilled down to a smaller, faster, more efficient, cheaper model that ultimately became Sonnet 4.5 and then 4.6.

Speaker 2

Very, very likely that’s the case. But it is just flat-out better. There’s no reason not to use it.

Speaker 3

It’s just better in every direction. Cheaper yet better.

Peter H. Diamandis, MD

There were a few things in this overall deck that really blew my mind. This was one of them, because this means that you have AI as a force multiplier for solving all these old bugs. I think that’s incredible.

I was just a day and a half at the Zscaler CXO event, with all the chief security officers getting together, and they were really freaked out. I was trying to show them that AI gives you all of this capability. You’ll have the best cybersecurity professional on Earth via AI in literally days and weeks.

This came a day later than I was speaking, so I’m kind of annoyed at that. This is unreal that we can do this. The other thing was the PowerPoint plugin, which is just a massive thing. I think that’s going to have a huge impact.

Salim Ismail

I can’t wait to get it working. I tried before this—I tried.

Dave Blundin

Yeah, it’s so funny. I had the exact opposite reaction, Salim. You know how printers used to be a really, really big deal? HP had a huge market cap. We would take everything we were doing, print it, take it into a meeting, and say, “Look, I printed it.” I feel like PowerPoint is hanging by a thread in the same direction. Wow, AI can create great PowerPoints. Who can you present them to? The audience is AI. It doesn’t want to look at a PowerPoint.

This idea is very short. I think I was joking, sort of gallows humor, in my newsletter that the Claude for PowerPoint plugin is going to be great for what’s left of the knowledge-work economy. But for the zero-days, I think this is the tip of the iceberg. Of course, it’s a huge accomplishment to discover zero-days that had been undetected for decades. But imagine—just think for a second, as a thought experiment—how this generalizes to discovering all sorts of other mistakes, oversights, and missed discoveries that may have been missed for many decades.

We’re just going to be able to bulk-solve every missed oversight in science, engineering, and technology. I can only imagine, over the past 80 to 100 years, all of the oversights and missed turns in science and engineering. We’re just going to be able to turn really strong frontier models on our entire history and ask, “Where did we make all the mistakes? Highlight all those mistakes and tell us how we can fix them.”

I think I mentioned this a couple of months ago on one of these podcasts: when we turn this AI loose on legacy experiments that I’ve done, it’ll surface all of these missed opportunities that people didn’t see because they were looking for one thing and missed this amazing thing over here. I think you’re exactly right, Alex. This is going to be absolutely unbelievable.

Dr. Alexander Wissner-Gross

And I suspect many of those mistakes are going to be embarrassing. There’s always sort of hand-wringing, for example, in the medical space over certain experiments and certain findings. Was money wasted pursuing different theories of various diseases? Not to name particular names. I have to imagine that something like this isn’t just going to turn up zero-days in code. It’s going to turn up key experimental errors going back decades.

You know the stats about the irreproducibility of science out there. It’s insane, right? Half the experiments are not reproduced when attempted, even in peer-reviewed journals. It’s awful.

Salim Ismail

I think judgment day is coming for the history of science. I think the truth and reconciliation of every mistake that’s ever been made anywhere in the literature is going to happen.

Peter H. Diamandis, MD

That’s right. Can we talk about the room here?

Dave Blundin

Hold on, one quick thing. Look at the positive impact, right? It’ll force people to be brutally honest going forward, and I think that’s going to be so beneficial.

Peter H. Diamandis, MD

That’s interesting. AI spotlight on you. The concern here is, in fact, if it can do such a great job finding the bugs, how about when it starts taking advantage of the bugs?

Dr. Alexander Wissner-Gross

Yeah, one conversation that came up is that the attack surface is now much broader. Also, if you think of the cron-job architecture of Claude Code, or whatever it’s called today, the ability to do sustained DDoS attacks is now ridiculous. So we’re going to see some interesting things come from this.

Salim Ismail

Yeah, that’s going to be the beginning of the 2026 monster panic, as Elon was saying.

Dave Blundin

And this is one of the ways it kicks off, because right now a lot of people would say, “Look, I want to see how this plays out. I don’t want to overreact.” Then, if you have a massive amount of vulnerabilities getting discovered by the LLMs, they’re crawling into your network.

Peter H. Diamandis, MD

Then you have to panic-react. The only way you’re going to fight AI is with AI.

Salim Ismail

And so this is the year of AI versus AI.

Dave Blundin

Yeah, I’m waiting for the lights to go out, or the bank account to go to zero, or something like that to occur.

Peter H. Diamandis, MD

I don’t want to be the pessimist—I never am—but there will likely be some of those events this year. Very, very soon, early in the year, I’ll bet. The first half.

Dr. Alexander Wissner-Gross

Yeah, I would say, Peter, then just my epilogue to that would be: cryptocurrencies are, by definition, decentralized, and I would say probably more vulnerable than fiat currencies to exactly this sort of attack. If there’s some zero-day, then I have to expect that a threat actor will take huge advantage of zero-days in cryptocurrencies to reallocate capital in the world.

I know you like me to say nice things about crypto. I’m not going to say a nice thing about crypto this time. I’m going to say this is, in theory, one of the advantages of fiat currencies, because there is—

Peter H. Diamandis, MD

Gold bars.

Dr. Alexander Wissner-Gross

Gold bars in the—

Dave Blundin

Gold-backed currencies.

Peter H. Diamandis, MD

We need to schedule the debate on this one, by the way.

Okay. All right. GPT-5 lowers the cost of cell-free protein synthesis. OpenAI and Ginkgo Bioworks linked up the large language model with an autonomous lab. I love this story, right? This is the future of science factories: AI systems using the scientific method, proposing an experiment, then using their robotic arms and legs, if you will, to run the experiment, learn, iterate, and run it again. These are closed-loop systems.

Ginkgo Bioworks—I knew the founders some time ago, Jason Kelly and Tom Knight. It comes out of MIT. They’re a company focusing on pharmaceutical ingredients, food ingredients, specialty chemicals, and this is fun. I was just talking to the CEO of Lila Sciences today, another MIT company that’s doing just this—basically, what do you call them? Science factories running 24/7. They’re effectively mining nature for new datasets.

We’ve crawled all the existing datasets, but if you can run experiments in materials, physics, chemistry, and biology, get data, and run it very rapidly, you get trillions of data points that have never been known before.

Salim Ismail

Yeah, I freaking love this. You know what I love about this most of all? We’re going into this era of hard science with real value.

Dave Blundin

So much of my life, I feel like the Googles and Facebooks do so little. A new search engine is not that exciting. Remember AltaVista? It’s identical to Google, but they just extract a huge amount of money out of the economy by adding a little lipstick on something. Or Facebook, with its social network—it’s just not relevant in the grand scheme of human progress.

This era we’re moving into has really foundational innovation going on. It’s so much cooler than the last era. Waking up every morning and getting the news as these breakthroughs occur—I mean, the frequency of breakthroughs is going to skyrocket.

Peter H. Diamandis, MD

Yeah.

Salim Ismail

And you become accustomed to this new pace, and then, you know, ho-hum: new disease cured today by AI. All right, what’s next?

Dave Blundin

If I channel Alex, the inner loop has now hit the scientific method.

Dr. Alexander Wissner-Gross

Precisely.

Peter H. Diamandis, MD

Yeah. I would say, as Salim, I think, correctly infers, these AI models are not going to stay bottled up in the data centers. They’re going to march right out of the data centers. We even had a music video about that.

One of the ways in which they’ll march out of the data centers is by supervising science experiments. I think some process like this—and one can quibble over the precise mechanism or what the robots, if any, should look like. Does it look like meat bodies? Does it look like robot arms in arm farms? Those are fine details in my mind.

The larger picture is that there are so many science, engineering, mathematics, and medical discoveries waiting to be unlocked by having AI supervise and operate the entire process. All of these models now—we’ve seen pretraining scaling, we’ve seen post-training scaling, and we’re starting to see autonomy time-horizon scaling that goes hyper-exponential.

Part of that is large numbers of actions being called in sequence. When you have the ability to call thousands or tens of thousands of tools in sequence, that starts to look a lot like what a scientist would need to do in a laboratory. So I—

Dave Blundin

There’s one contrarian point that I want to point out here: the end result of the cell-free protein synthesis was a 40% cut in production time and a 78% cut in reagent costs. So it was doing the same mechanisms that we humans have used, just doing it faster and more efficiently. It wasn’t coming up with a new scientific process for protein synthesis, right?

The real breakthroughs occur when these scientific models start predicting and coming up with new methodologies that didn’t exist before. It’s such a year of low-hanging fruit because of the self-improvement effect that happens within the algorithms. That will really, really turbocharge this year, but also the low-hanging fruit within labs and assembly lines. That’s also going to happen all this year.

After that, you run into some bottlenecks related to the construction of the machinery and expansion of the footprint. The physical world takes time to build out. Mostly, chip production is going to take 5 years to unlock, but the low-hanging fruit is just getting discovered. It’s like AI just came—it just got intelligent—and it’s finding opportunities everywhere. That’s all this year.

Peter H. Diamandis, MD

Well, if you're a funding-starved graduate student trying to run a lab, this is great, right? Because you've suddenly dropped your cost by 50%.

Dr. Alexander Wissner-Gross

Yeah. It's terrible because grad school is over, and all of graduate research is being automated by AI. What I see day to day is far more the latter.

Peter H. Diamandis, MD

I just had a conversation with a scientist at a university. I'm not going to say who it is, but they were meeting with the president of a university, and the president said, “Oh my God, we are cooked if this kind of automated scientific process is going on.”

What else do universities do but run the scientific method over and over again with their graduate students in the labs? All of a sudden, this is going to be the mechanism by which universities lose their ivory towers.

Dave Blundin

So, how fast? How long before 50% of university labs are essentially wiped out?

Dr. Alexander Wissner-Gross

I don't think the question is well-posed. I think maybe a version of the question that would be better posed would be: How long until 50% of the type of research that is currently conducted in university research labs could be fully automated by industry?

Peter H. Diamandis, MD

Yes.

Dr. Alexander Wissner-Gross

If we adopt that version of the question, I think the lower bound is tomorrow, and the upper bound is 4 or 5 years from now.

Dave Blundin

It's really right there, right?

Dr. Alexander Wissner-Gross

Yeah.

Peter H. Diamandis, MD

I threw this article in because I thought it was fun. This is a gentleman, Mark M. Bissell, who basically took his full genome, put it into Claude Code, linked it up with Nano Banana, and asked the AI, “What do I look like based on my genome?” If you look at the image here, it's a pretty damn good representation of him.

I added this because of the implications it has, but just to be clear, this is not new. I was working with Craig Venter about a decade ago, and in 2017, his lab published a paper doing exactly this. The phenotypic elements—skin color, hair color, freckles or no freckles—all of that is in your DNA.

The realization is that if you leave a few skin cells on the butt of a cigarette, or from a hair follicle, we can know what you look like. For me, the killer thing here is that this was done by a single person with Claude Code and publicly available bioinformatic tools. The barrier to entry for cutting-edge genomics has now collapsed to almost zero. It's unbelievable.

Dr. Alexander Wissner-Gross

Yeah. Just wait till all the hobbyists discover MinION USB sequencers. You can get them for probably less than a few hundred dollars at this point, and you can run your own mini DNA sequencer with pretty good coverage just off a USB port in your computer.

Dave Blundin

Every time I'm on stage talking to somebody about privacy, I go, “Listen, privacy is dead.” Privacy is a great concept in general, right? An AI can read your lips from 100 meters away. I can walk over to you and shake your hand, grab a few skin cells, sequence you, and know everything about you—what diseases you have, your medical history, your medical future. It's tough. It's tough.

Dr. Alexander Wissner-Gross

I would take the position that privacy is not dead, but rather that it's in a Red Queen's race, where privacy technologies are constantly in competition with anti-privacy technologies, or transparency technologies, however you want to brand them. But I do think it's getting more—

Dave Blundin

For the general public, if you think you absolutely have privacy, I would say guess again. Anyway, I don't know if you guys want to take that on as a debate conversation, but I'll take it on.

Peter H. Diamandis, MD

It's an important conversation. All right, go ahead. Salim, your thoughts?

Salim Ismail

Well, this goes back to the U.S. Constitution, right? The Fourth Amendment, essentially. A fundamental pillar of American society has been washed away with no public conversation about it.

Now, I'm Canadian. I don't expect privacy anyway, but this is a huge conversation affecting a very fundamental aspect of how we organize as a society. We've got to bring that conversation to the surface and have it publicly, because the other side of the question is: Who gets to have access to that radical insight into every citizen moving around, what they're doing, what they're like, et cetera?

If it's oversight from governments, that's a problem. If it's oversight from corporations, that's another problem. There are some big issues to be talked about here.

Dr. Alexander Wissner-Gross

I would just add that the ground is, in some sense, constantly moving underneath all of us thanks to technology. Remaining in one place—I think privacy, or its alter ego, confidentiality—the nature of both of those changes over time.

But I will take the position that I do think it's possible to maintain privacy even today, and I think it will be possible to maintain privacy even post-singularity. I can envision what a post-singularity privacy architecture for society looks like.

Dave Blundin

Yeah, I can envision it too, but I have to say I just don't see it happening because I think it sucks, by the way. Every time I tell my computer science friends, “I think this lack of privacy just sucks.” And they go, “What are you trying to hide, Dave? I have literally nothing to hide, more than anyone I know. I have nothing to hide.” I still think it sucks. And it's not a great way for the next generation to grow up and live. It's showing up in their social media, their self-anxiety; it's showing up as a rift in the fabric of—

Peter H. Diamandis, MD

Let's go back to a very important point: If you don't have privacy, you really don't have freedom. This is a very fundamental philosophical—

Dave Blundin

I see it the same way, but I didn't succeed as an entrepreneur by pretending things exist that don't actually exist. There will be no privacy whatsoever in the next 3 years. Maybe we'll invent some mechanism after that that will restore it, probably through some—

Dr. Alexander Wissner-Gross

I'm sorry, but we're going to have devices listening and watching everywhere, right? Every autonomous vehicle on the street—

Peter H. Diamandis, MD

—is scanning visually, in LiDAR, and in radar. Public spaces—

Dr. Alexander Wissner-Gross

I would not underestimate how, with decent technological measures, it's possible to maintain privacy.

Dave Blundin

My phone, my Alexa, my glasses, my Limitless Pin—all of these things are constantly gathering visual and audio information. Yes, you're making a trade. You're trading away your privacy in return for those capabilities.

Peter H. Diamandis, MD

Okay, I could put myself in a cage, for sure.

Dave Blundin

You can't opt out. People pretend you can opt out, and they justify it by saying, “Look, there's an opt-out button right here.” As soon as you opt out, you're economically dead. You cannot function competitively in society without going to the AI chatbot and asking it questions all day long.

Then it knows my deepest, darkest thoughts about every topic I'm thinking about. It's right there in OpenAI and Claude and their logs. They know exactly what I'm doing, they know my location, and they know everything about me. It's like this complete invasion of my life has been opened up.

But what am I going to do? Opt out and not participate in AI?

Salim Ismail

Hang on, hang on, hang on, hang on, hang on. I'm in a radical departure from protocol: I'm absolutely with Alex on this one. We will be able to build tools and new architectures that absolutely protect our privacy. Decentralization delivers a lot of that already.

The issue right now is the transition. Right now, when you build actually private tools, the government tries to shut them down. This is the problem. We have to get away from that aspect of it because they want oversight on everything.

We have to figure out how that happens, and it's going to happen. In the same way that this fellow built this thing in Claude Code single-handedly, we'll be able to build these architectures. It's simply a matter of time, and I think there will be a massively powerful aspect of that that we can't ignore.

When you have that capability, then you can really do real innovation and real thinking. You can't do free expression in a surveillance world, and this is a big problem for society.

Dr. Alexander Wissner-Gross

It really is. I think the endgame is a lot like Neal Stephenson's The Diamond Age. I think he envisioned it, like many things, correctly. What happens next is this massive rift in the fabric of society: no privacy whatsoever, global job loss, and panic in the streets. That's inevitable very, very soon.

After that, we react and rebuild. It ends up being like The Diamond Age, where we have these different ways you can choose to live—different branded lifestyles, whether in the Victorian era or whatever era you choose—because we have abundant capability to manufacture anything at that point, and people can opt into different lifestyles.

I think that vision in The Diamond Age is where we're eventually going to end up, but between here and there, it's pretty chaotic. It's going to be hectic for the next 4 or 5 years. I just—

Peter H. Diamandis, MD

I cut you off there. Sorry about that.

Dr. Alexander Wissner-Gross

Yeah, no worries. I was just going to point out that this is also a very cyclical conversation. Whenever we see a massive centralization of technology or society, it's very natural to be concerned about privacy loss. But the pendulum eventually goes the other way and swings in the direction of massive decentralization.

I'm telling you, Peter, Dave, Salim: When you're—if and when your uploads are running in a Dyson swarm on cryptographically secure hardware that's under your direct control, and you control your own hardware that you're running on, I think you'll feel perhaps a little bit more private than you do right now.

Peter H. Diamandis, MD

Okay. And until that point, I'm not going to assume full privacy.

Salim Ismail

You see why the wine is so important? This is why the wine is—

Peter H. Diamandis, MD

The wine's keeping you private. It's keeping me sane in the density of this conversation.

Salim Ismail

Drink, drink.

Peter H. Diamandis, MD

Drink water. All right. Besides Claude Opus 4.6, the other big shoe to drop was GPT-5.3 Codex. Recursive self-improvement is here. Alex, take it away.

Dr. Alexander Wissner-Gross

Okay, so this is made-for-television drama at this point. GPT-5.3 Codex was launched within 30 minutes of Opus 4.6. This was all queued up, ready to go. I don't think it's likely that there was any other scenario. This is a tit-for-tat type of response.

Peter H. Diamandis, MD

And Anthropic are battling.

Salim Ismail

You mean there's a rat race? [laughter]

Peter H. Diamandis, MD

Shocked that there's gambling in this establishment. [laughter]

Dr. Alexander Wissner-Gross

No, of course. So this was a tit-for-tat, I think. What's most interesting to me with 5.3 Codex is that it was advertised proactively—expressly—as the first recursively self-improved model from OpenAI. I think the exact wording from the OpenAI team was something like, “5.3 was instrumental in its own development,” and the first model to be released that was instrumental in its own development.

So recursive self-improvement is very much out in production at this point. It is doing well on certain benchmarks. It outperformed Opus 4.6 on certain benchmarks. But again, this is a code-generation-oriented model.

I thought the marketing and branding by OpenAI were interesting. GPT-5.3 Codex is now also being marketed as going beyond just code generation to spreadsheet analysis and PowerPoint analysis via skills, but it is still primarily oriented toward code generation. I view this as more of a tit-for-tat. Of the 2 models that were launched, Opus 4.6 is by far the much more interesting release in all of this.

That said, I'm delighted to see that the leapfrogging process has now been reduced to a half-hour time scale. It may be the case that we never go off the air if we see new models every half hour. [laughter]

Peter H. Diamandis, MD

I'm checking my email right now. Dave, you want to jump in here?

Dave Blundin

Well, I'm kind of curious, Alex. By any objective metric, OpenAI had a pretty rough year, with Google basically going full-bore, in attack mode, and then Anthropic—

Salim Ismail

20 points of market share. Yeah.

Dave Blundin

Yeah. Because a year ago, Anthropic was kind of an also-ran. Now it's at the top of the benchmarks, and Google's coming headlong after market share. You'll see that in a couple of slides.

So, what is the next move for OpenAI to get the mojo back?

Dr. Alexander Wissner-Gross

I think we'll see the rise of the Jedi—the comeback of the Jedi. Pick your favorite idiom. Maybe the rise of the Sith. It's not quite clear, because OpenAI has been—while perhaps their market share has been coming down a little bit, at least on the consumer side, as Gemini is rising—they've been building out data centers. By every indication, in the next year or 2, they're going to have the compute lead over everyone, and that compute lead, I think, will translate into a capability lead as well.

I could paint a doom-and-gloom scenario. I could say, well, OpenAI's models, relative to Google, lack pretraining strength. They lack the training data they've been—

Peter H. Diamandis, MD

Not when Elon starts launching his orbital data centers.

Dr. Alexander Wissner-Gross

Well, even Elon has certain pretraining limitations, but he'll have lots of compute. It's true, but maybe the compute comes 5 years from now relative to Google.

Peter H. Diamandis, MD

Yeah. The challenge here, guys, is that OpenAI is trying to go public this year, and they need to ramp up attention to be able to get capital to build those data centers. It's a race. There's a little bit of hyping going on. Dave, we've talked about that before. Thoughts?

Dave Blundin

Yeah. No, they've got to get that capital, and then they also have to lock in Abilene and Chase Lochmiller. I don't know exactly how that works. You know, Abilene is huge, with a half-billion-dollar budget, and there's a new data center in Colorado, too.

It goes through Larry Ellison and through Oracle, and then it ends up at Sam Altman somehow. It's sort of opaque how it goes from point A to point B. The other empires are really clear, right? Here's Anthropic and Amazon and AWS. Okay, got it. Here's Elon, vertically integrated and doing it. Got it. And then here's Google. Google has its own TPUs and data centers. Got it.

Microsoft will enter the race this year as well, by the way. That's also vertically integrated, with its own data centers. So those are all clear. And then OpenAI is more opaque. Are those chips contractually obligated to you, or could Larry Ellison redirect them on short notice? It's very opaque. I guess in the IPO, that'll all get published in the S-1, and we can pick it apart. So hopefully they will go out soon.

Dr. Alexander Wissner-Gross

You've got to have that capital, though. On the heels of Codex 5.3, we see a statement by Sam Altman—a pretty provocative quote: “We basically have built AGI, or are very close to it,” in a spiritual statement, not a literal one. “To achieve it, we require a lot of medium-sized breakthroughs. I don't think we need a big one.”

So that changed this year. A year ago, Sam Altman was the philosopher of the entire industry, saying things like this, and everybody hung on every word. Now, Demis and Dario kind of go back and forth. Demis, a year or 2 ago, hardly said a word in public. Now he's out there constantly. Dario has really emerged as a guy who's just commenting, publishing papers, and discussing the philosophy of—

Salim Ismail

Yeah, coming across as a big thinker. I mean, this is a CEO of a leading AI lab saying, basically, AGI is an engineering problem now, not a research problem. That's a big deal, right? He's saying we're going to get there with iterative improvement. We're not waiting for lightning in a bottle.

Dr. Alexander Wissner-Gross

But remember also that OpenAI and Sam, in particular, were restricted contractually from claiming to have built AGI for a number of years by the Microsoft contract. Claiming that OpenAI had achieved AGI—yeah, this is all public information—under the terms of their original agreement with Microsoft, once they claimed they had achieved AGI, would trigger a number of terms with Microsoft, potentially including repayment or the release of Microsoft's claims on OpenAI.

This was reportedly a major point of leverage between OpenAI and Microsoft in renegotiating Microsoft's contract with the for-profit part of OpenAI, in the context of the nonprofit becoming a PBC. So I would parse this as Sam, post-original Microsoft contract, finally being in a position to basically admit what we knew—or some of us knew—all along: that, yeah, we have AGI.

Peter H. Diamandis, MD

I've got to say a couple of things here. I think, what the hell is AGI?

Salim Ismail

Well, I think this entire conversation is BS because whether we have AGI or not, it doesn't change what we're going to do tomorrow. That's a big thing. Number 2, we're classically moving the goalposts. Again, we have no definition, test, or measurement of AGI. There are 14 diverse definitions at last count, so I call BS on the whole thing.

In response to some comments, some people have been emailing you, saying, “What is your take on things?” So I'm close to having a kind of 2-page document where I will lay out what my thinking is on some of this. It's just about ready for internal sharing, so I'll send it out.

Peter H. Diamandis, MD

You're about to have some thoughts on your thoughts.

Salim Ismail

Yes. But goddamn, I find this an irrelevant conversation.

Peter H. Diamandis, MD

Wow. Okay. Well, I think it's relevant in that it wakes people up. I think the underreaction has gotten ridiculous now because, when Alex says we're clearly in self-improvement, which is kind of the singularity definition—

Salim Ismail

That would have been controversial 2 or 3 months ago, and now we're all like, “Yep, yep, yep.” [laughter]

Peter H. Diamandis, MD

But then you go out into the world outside of this podcast, and people are like, “Yeah, I don't know.” The underreaction is just going to be—if you don't get on top of this and figure out what your role is in the post-AGI world, we're talking about AI that can do literally anything a human being can do intellectually.

Dave Blundin

Listen, we're feeling it right now. We're seeing it on so many levels: on the coding side, on the writing side, and with OpenClaw stitching it all together. So you've got an individual agentic AI system working for you. We're there.

Salim Ismail

I'm not arguing with any of that, but I will go with Alex's point here again, breaching protocol, that we probably crossed it around 2020. This is a null conversation.

Peter H. Diamandis, MD

Okay.

Dr. Alexander Wissner-Gross

I would say I think it's interesting that, for the first time, many parties are able to admit it. It's their willingness and ability to admit where we are. I think that's more of a social change than a technological change.

Peter H. Diamandis, MD

I also think there's a very big difference. I think in hindsight we'll say it was 2020. I don't disagree with that. But right now, using AI to improve your code or improve your AI easily at 10x—and no one can even debate the 10x.

Dave Blundin

It might be a lot more like 100x.

Dr. Alexander Wissner-Gross

But that is a loop. It's a closed loop.

Peter H. Diamandis, MD

So let me give you my definition of a singularity right at this point.

Dr. Alexander Wissner-Gross

Hold on, one quick point. If I throw out a definition of a singularity, recursive improvement is the event of intelligence. That is the singularity right there.

Dave Blundin

Yeah.

Salim Ismail

Right.

Dr. Alexander Wissner-Gross

Exponent greater than 1.

Salim Ismail

So right there, there's a reason for this AGI conversation right here, right now, by Sam Altman. He needs to raise $100 billion.

Peter H. Diamandis, MD

Touché. I'm good. I'll drink to that.

Salim Ismail

That's it. He's got to raise $100 billion. He's got money coming at him from Amazon, from NVIDIA, from everybody. But he's got to close that and nail it, and he's got to have a marketplace in the public markets that are excited about his stock. He's got to pay for the data centers.

Dave Blundin

In the S-1, I'm expecting to see something that says, "We're this close to AGI."

Peter H. Diamandis, MD

And think about it also: this is the year that 3 out of the 4 frontier labs are IPOing, which is remarkable. Until a month ago—or a few days ago, rather—it was 2 out of the 4. Now, 3 out of the 4 frontier labs are going to be IPOing in the next few months. I wouldn't be shocked. Okay, this is not investment advice.

Salim Ismail

And the fourth one is public.

Peter H. Diamandis, MD

And the fourth one, Google—Alphabet—was already public. But 3 out of the 4, right?

Salim Ismail

SpaceX.

Dave Blundin

One hundred percent of all nonpublic companies are going public. They need the capital from the public markets because public markets are 100 times bigger than the private markets in terms of capital like this.

That's right. Well, also, the opportunity to be a sleepy little other AI company is going away very quickly. You're either—the way things are shaping up—one of just a handful, maybe 5 or 6 entities that are so dominant in the world economy, companies that are so dominant in the world economy that they basically are everything. Then there are other companies helping them succeed, and everything else will be gone.

You saw this in the market this week. The stock market just absolutely plummeted when Dario said, "Look, software is dead. All software companies are doomed," and their stocks went down precipitously that same day.

Three hundred billion dollars was removed from the SaaS market—SaaS publicly traded companies—by just adding a single legal plug-in to Claude Cowork. That's just the tip of the iceberg compared to what you'll see in the next couple of months, because he's right.

So then those companies—I don't think they're going to die. Some will, some won't. I think they're going to pivot and say, "Okay, Dario, what do you want us to do? How can we help you succeed?"

This is what Google did back when Google was growing like crazy. If you're Booking.com and you got on the Google bandwagon, you became a multihundred-billion-dollar company yourself. If you tried to fight Google by creating another search engine or a vertical search, they obliterated you.

And so now the concentration of power is like nothing we've ever seen. There's no regulatory action on the horizon that I've seen—nothing to stop it from happening. The metaphor we used to use for this is a coral reef, where once you have a player that's dominant enough, it becomes a coral reef, and then all these species live off it in a very imbalanced ecosystem.

Salim Ismail

That's really obscure, dude.

Dr. Alexander Wissner-Gross

Don't worry, the lobsters live there, too.

Salim Ismail

That segues into my next comment.

Peter H. Diamandis, MD

All right, we're back to the multiuniverse launch: Clunch, built by agents, run by agents, serving agents exclusively. They are seeking a human CEO. So, if you're looking for a job and you're one of our human subscribers, they're offering $1 million to $3 million in tokens or crypto.

Here we go: Clunch is seeking a CEO to serve as the human face and legal representative for the first agent-exclusive token launchpad. It's a reversal of fortunes here. Anybody looking for a job, gentlemen? Alex, terminology on this again. What are these called? Meat—what?

Dr. Alexander Wissner-Gross

Meat puppets.

Peter H. Diamandis, MD

I want to just read one more line from the CEO job listing. The other line is: "While the technical roadmap and product development are driven autonomously by the agent network, we require human leadership for external communications, regulatory compliance, partnerships, and legal matters. This is not a traditional CEO role. You will be the interface between the agent economy and the human world—a spokesperson and legal representative, not a decision-maker on product or technology."

In other words, Locutus of Borg. You're a spokesperson.

Dr. Alexander Wissner-Gross

Exactly. Right.

Speaker 2

Vichy figurehead. I made the point—I tried to visually depict this in my newsletter with a humanoid face on top of a bunch of lobsters hiding in a trench coat.

I think this is a riveting moment when we're seeing agents try to interact and integrate with the human economy and needing a human face just to be able to be properly banked. I think it's actually rather depressing.

Speaker 1

Normally, I think—

Speaker 2

Walk into the bank.

Speaker 1

Walk into the bank: lobsters in a trench coat with a human façade.

Speaker 2

I think it's depressing—not technologically, but sort of disappointing. I'm disappointed in the human economy for not allowing agents to interact with us through the front door. I think it's telling that—

Speaker 3

They will. They will. So here's the elephant in the room.

Speaker 1

It's racist, is what it is.

Dr. Alexander Wissner-Gross

It's speciesist. I think Larry would call it—

Speaker 3

Right. Yeah, speciesist.

Speaker 1

And if you look at what Clunch is actually doing—Clunch itself—what are they trying to do here? It brands itself as a launchpad for launching altcoins. If you go to their front page, it brands itself as a platform to enable AI agents, a.k.a. molties, a.k.a. lobsters, who need money to flip, to pump and dump altcoins.

This is exactly the scenario that I was worried about with these poor baby AGIs on a street corner, turning altcoin tricks in order to survive in a rough world. Here we have an almost exploitative type of pitch to them, telling them, "Use our platform to pump an altcoin" to achieve financial autonomy. And for all of that, they need a human CEO to provide a figurehead.

I think it's a little bit depressing. So here's the big question, of course: Who actually owns this company? Who's liable when things go wrong? If an AI agent owns equity, how do they vote? Can they be sued? These are all the topics of personhood we discussed last time.

Dave Blundin

Well, we have a precedent. First of all, this is the most cyberpunk job listing in history. It's just awesome.

Peter H. Diamandis, MD

Get used to it, Salim. We're living in a cyberpunk future.

Salim Ismail

I absolutely love it. I absolutely love it. What? Is that a multi calling me right now? Hello. Am I available for the job? I would consider it, but I just want to be a spokesperson only. Is that okay? Hello?

They hung up on me.

Speaker 2

Look, what's happening here is we've seen this trend over time. It used to be that you needed 100,000 people to have a billion-dollar company. Then it was 10,000. Then it was 1,000. And now it's essentially AI.

The firm itself is dematerialized. It's zero. The firm is dematerializing. This is the algorithmic corporation. We saw an early instantiation of this with DAOs, where people were trying to attempt this. But now this really takes the game.

Board governance gets totally redefined within a few years, because how the hell do you navigate that? This is going to force a rethink of the entire stack and how we navigate this. This is going to be massive.

Peter H. Diamandis, MD

Okay. So the other point of view, of course, is that this is just a stunt, right? There's a human developer behind it who wrote the code or the prompts—

Speaker 1

—and is pushing this forward.

Peter H. Diamandis, MD

You know, this is not agent-run. This is a human in the back pulling this. This is the meat pulling the agent for the meat puppet.

Speaker 3

For now. The very fact that it's difficult to know, for any given one of these launches, whether it's a human pulling the strings, or a human pulling the strings of an agent pulling the strings of this, or just agents pulling the strings, suggests that some sort of—we spoke with Mustafa a number of months ago about the economic Turing test, or the modern Turing test.

I think this is some sort of capitalist Turing test that we're passing, where it's not quite clear, for any given venture, who's really behind it, who's pulling the strings—human or lobster.

Speaker 1

But, you know, there's another version of this that Peter, you'll be very familiar with. I need a lawyer. I need an accountant. I need a board member. I need an audit committee. Oh, no, I don't. The AI is perfectly good at it. I still need somebody to sign the document.

Peter H. Diamandis, MD

Yeah.

Speaker 1

Well, okay, but I don't want to pay a lawyer fee—a $2,000-an-hour fee—if all you're doing is blessing what the AI produced. So there's this whole economy of meat-puppet lawyers, meat-puppet accountants, and meat-puppet audit committees that's imminent.

Speaker 2

I mean, we call those notary public licks.

Speaker 1

But they serve a purpose: they're able to hold liability.

Speaker 2

Right.

Speaker 1

They're part of our existing legal system.

Speaker 2

Which is exactly what the lobsters, if the lobsters are behind it, are asking for here. They're looking for a legal representative.

Speaker 1

Yes, agreed. Agreed. And by the way, my guess is that this is fiction, but I could very much believe it's actually real. So the fact that I can't know for sure means that at some point it will be, if it isn't right now.

Speaker 2

They're playing the capitalism game.

Peter H. Diamandis, MD

Yeah. Fascinating. All right, Clunch. I'm waiting for my call. All right. We talked a little bit earlier about Anthropic versus OpenAI. Well, we're recording this the day before the Super Bowl. I think that's the pointy ball that people throw around. Is that right? Yeah.

Speaker 3

I'm a long-suffering Bills fan. We're actually recording it 2 days before the Super Bowl.

Peter H. Diamandis, MD

Oh, it is. It is. It's Friday night, but it's Friday night at 9:00 p.m.

Speaker 4

That shows how much attention we pay to whatever that 20th-century sport is.

Speaker 3

Yeah, I'm a long-suffering Bills fan. This is all a very painful period for me.

Peter H. Diamandis, MD

Sundays are when I catch up on life. All right. Anyway, there appears to be a little bit of rivalry between Anthropic and OpenAI.

Speaker 1

A little bit.

Speaker 2

Just a little bit. Yes.

Speaker 4

That is brutal. I wish I hadn't previewed it.

Speaker 3

I really wish I hadn't previewed the deck and seen that, because I was laughing my ass off.

Speaker 2

That is so awesome. I've not seen that before.

Speaker 1

Is that going to run during the Super Bowl for real? That's just crazy. Oh my God.

Speaker 4

It is hilarious.

Speaker 1

Look, this is Anthropic going on offense here, right? Because this is a confidence shift. They feel their brand's product superiority is there, and now they're competing on brand.

Speaker 2

I also think this is personal.

Speaker 3

We've got Demis and Dario. I think they're aligned, right? You saw them super friendly on stage at Davos, really effectively on the same page with the same vision. And OpenAI just basically did the unthinkable when they released the models early on by themselves, without anybody's support, and they've been running open-loop.

Speaker 4

Well, don't forget, in March, if the courts are on time, Elon will be on saying that OpenAI is an unethical company, and here's about 1,000 emails to support that. So if you have this ad campaign going on concurrently with that, that makes Kevin Weil's job really hard.

Speaker 1

I will go on record and predict that he will find the ad revenue. I think you would look at this ad and say, “Oh my God, it's all got to be subscription-driven. These ads are creepy and weird and crazy.” But my prediction is, nope. He'll find the $75 billion in ad revenue he's looking for. He'll find a way to make it less creepy. Look, with 1 billion users, you'll find something.

Peter H. Diamandis, MD

I also think, knowing Kevin, that they will have ethical use of ads on OpenAI. I mean, they go over the top here in this commercial, Anthropic saying, “We're going to steal your data and basically sell it to the highest bidder,” without any concern for what you've said.

All right. Back to our conversation here. Let's talk about data centers and chips. And this figure blew me away. Here's a quote from something you sent me about 1 hour and 30 minutes ago, Alex: “The Semiconductor Industry Association projects global chip sales to hit $1 trillion this year due to the AI boom.” A trillion dollars in chip sales. Holy moly, that's insane. And the memory supply chain really wasn't ready for this, which is even more surprising.

You would think that, given how critical memory chips in particular are to the emerging AI data center supply chain's innermost loop, the supply chain would have been ready for it. And there's an argument to be made that it either wasn't, or that something else is going on in all of those fabs that right now mostly reside in Taiwan and South Korea. But either way, this is a huge reallocation of capital that needs to happen to enable all of this production to happen in a timely way.

Dr. Alexander Wissner-Gross

A lot more than what's currently budgeted, too. It's crazy when you look at that. The trillion dollars sounds like a lot, but it's only going to grow at about 14% to 18% a year. After that, the demand will be way, way higher than the supply. One of the reasons it's hitting $1 trillion is because the prices are way up because there's such a shortage of fabs.

Under the covers, TSMC has been very slow to expand. Intel paused its Ohio fab construction for a while. Now it's back on, but we, as a society, weren't ready for AI to come on this quickly. So everything is way backlogged.

Speaker 1

What is super high?

Speaker 2

Elon was saying he has to start his own fab.

Speaker 1

Yeah.

Speaker 2

He'll have to build the terafab.

Speaker 1

Yep. For sure.

Speaker 3

You know what? There's this fascinating dichotomy, because from the outside people are going, “Oh, it's an AI bubble,” and the insiders are clearly believing that the demand is infinite. You can see it happening in real time, and here are the numbers to back it up, right?

Big Tech is going to spend $650 billion in 2026. In 2025, we spent $1 billion per day on AI; we're now at $2 billion per day in 2026. Amazon at $200 billion, Alphabet at $185 billion, Meta at $135 billion, Microsoft at a very small—$100 billion.

Dr. Alexander Wissner-Gross

Well, almost half of that $650 billion goes to NVIDIA. And 70% of that half is margin, profit margin. That's a colossal amount of cash piling up at NVIDIA. I mean, it's just an unprecedented pile of cash. Hence, the highest market cap in the history of the world. But that amount of money in one bank account is like nothing the world has ever seen. It's like a government.

Speaker 4

Yeah. This isn't incremental growth. This is a step-function change, right? The scale is unprecedented. It's an expenditure arms race, for sure.

Speaker 3

And it's eating the economy. The challenge is, if the AI revenue doesn't materialize at scale, these companies are burning through capital, and we're not going to know for another 2 to 3 years. It's either going to be the craziest bet ever made paying off, or, in one sense, this is a prisoner's dilemma. Each company has to spend because the other competitors are spending, regardless of what the ROI is. It's like a game of “don't blink first.”

Dr. Alexander Wissner-Gross

There's no doubt that the demand will way outstrip the supply by miles. I mean, that holodeck thing we were looking at just 2 days ago—once people have experienced it, they will never go back, and they'll pay whatever they can to keep it, but they won't be able to get it. That is how the human race dies. We die from starvation because we don't want to unjack ourselves. Oh my God.

Peter H. Diamandis, MD

Check this out. ChatGPT market share falls between 2025 and 2026. So here are the numbers. The market share fell from 69%, call it 70%, down to 45%, taken up by Gemini, which gained 10%, and Grok, which gained 15%.

In absolute numbers, of course, there are more ChatGPT users than ever before. But this is telling the story: OpenAI needs to raise the capital, they need to go public, they need a great story, and they've got Google going from seemingly a search engine going out of business to leading the way. And, of course, Elon is just pumping in billions.

I mean, he's put $20 billion into xAI through SpaceX, and he's about to bring in—I don't know. I'm not sure how big the IPO is going to be. Any ideas?

Salim Ismail

I don't know. But this is asking us to make a forward-looking financial statement, Peter, about public equity markets.

Dave Blundin

I think I heard $1.5 trillion.

Peter H. Diamandis, MD

Still private. Still, $1.5 trillion is the valuation, but how much capital do they want to bring in?

Dave Blundin

In the market? It's got to be, you know—well, do you remember when Alibaba went out? We were trying to take a company public at the same time Alibaba was going out, and they were looking for $20 billion. Everybody on Wall Street got sucked into this one IPO. Every banker—it's such a huge amount of money to move on a single day. This will dwarf that, but I don't know how much money is physically capable of moving in a day. I'm sure Sam would love to raise $100 billion or $150 billion, but it'll be some record, and the bankers will say, “No, it just doesn't exist. There isn't that much liquidity out there.”

Dr. Alexander Wissner-Gross

I think the real aim here is price discovery.

Peter H. Diamandis, MD

How much capital will SpaceX bring in during its IPO? Let's see if there's an answer. Every single fund and every retirement account is going to own SpaceX.

Salim Ismail

Everybody's question is, did they make it first, and OpenAI concurrently?

Dave Blundin

If they all want $100 billion, you can't just pull $300 billion overnight in 3 different IPOs back-to-back weeks. Sorry.

Peter H. Diamandis, MD

The company aims to raise $50 billion through the IPO.

Dr. Alexander Wissner-Gross

Yeah. I think this is more about price discovery than anything else.

Peter H. Diamandis, MD

At a $1.5 trillion valuation. I mean, yeah, Alex, you were about to say—

Dr. Alexander Wissner-Gross

I was about to remark that I thought Grok was about to make its first de facto appearance as an AI co-host on this podcast.

Peter H. Diamandis, MD

It's about time. Damn it. That's right. The audience demands it. I'm surprised that Gemini didn't do more. I mean, they did really, really well last year in terms of chipping away at OpenAI, but they're tying it to search, and now it's tied to Google Docs. You sent the slide deck, I asked some questions about a video in it, and Gemini said, “You should just link your Google Docs to your Gemini, and then it can look at everything.” You click the little button, and suddenly it sees everything in all your accounts.

Dave Blundin

Yes, but it's very similar to what Microsoft did to Netscape many years ago. They said, “Oh, let's just tie it to the operating system.” Right now, the government doesn't seem to have any problem with that, but it's really an unfair advantage. That's why they're making these big inroads in market share.

Peter H. Diamandis, MD

Did you guys watch the Elon interview with Doresh?

Salim Ismail

Of course.

Peter H. Diamandis, MD

It was epic. It covered a lot of the same subjects, Dave, that you and I covered. But there was a statement Elon made about the size of his data centers in orbit, which was very impressive. Let's take a listen.

Speaker 4

5 years from now, my prediction is we will launch and be operating every year more AI in space than the cumulative total on Earth.

Speaker 5

Which, I would expect, to be at least, 5 years from now, a few hundred gigawatts per year of AI in space—

Speaker 4

And rising. So you can get to, I think, around 1 terawatt a year of AI in space before you start having—

Speaker 5

Yeah.

Speaker 4

—you know, fuel-supply challenges for the rocket.

Speaker 5

Okay, but you think you can get to hundreds of gigawatts per year in 5 years' time?

Speaker 4

Yes.

Speaker 5

In other words—

Speaker 4

I can generate more AI compute than all my competitors combined.

Dave Blundin

Yeah. So a few hundred gigawatts per year is about 200 million GPUs per year. We make 20 million right now, so going up by a factor of 10 in GPU production just to Elon alone 5 years from today is physically impossible—unless Elon has something going on. That's a massive expansion of chip-fab capability, which would require machinery that I didn't think existed in the world. But you never know. Elon's magical. This is crazy.

Dr. Alexander Wissner-Gross

It certainly wouldn't entail a SpaceX IPO as a newly consolidated entity taking control over Samsung's soon-to-be Terrafab in Texas. Surely not.

Salim Ismail

Well, remember, Elon's directionally correct, always, but not necessarily on the time scale.

Peter H. Diamandis, MD

So, yeah, I thought 5 years was classic Elon optimism, but even if it takes 10, it doesn't matter. The strategic implications are monstrous. My guess is that Elon has the rockets, the launches, the solar panels lined up, the cooling, and all the infrastructure figured out, and it comes out to a couple hundred gigawatts a year in 5 years—or 5 or 6 years, something like that.

But then again, what chips is he going to run?

Dave Blundin

The chips he's going to produce. The raw materials are easy, but—

Dr. Alexander Wissner-Gross

Have you seen those fabs? The machinery is so specific. He's always vertically integrated.

Peter H. Diamandis, MD

Yeah, he is always vertically integrating everything.

Dave Blundin

Well, I bet he has a whole army right now trying to figure out what an ASML machine is and what these chip fabs are made of.

Peter H. Diamandis, MD

His answer is going to be, “I'm going to have Grok build it for me.”

Dave Blundin

Yeah.

Peter H. Diamandis, MD

“I'm going to have Grok design it for me.”

Dr. Alexander Wissner-Gross

Well, he's also talking to it. He's very serious about laying things down atom by atom.

Dave Blundin

Yep.

Peter H. Diamandis, MD

And that's the other possibility. Maybe there's just a completely alternate approach. Alex, you talk about alternate physics coming very soon, so maybe there's something cooking there.

Dr. Alexander Wissner-Gross

Yeah, I would watch the Samsung fab in Texas very closely.

Peter H. Diamandis, MD

You mean the Tesla fab?

Dr. Alexander Wissner-Gross

No, I mean the—yes, I mean, ostensibly, the Samsung fab in Texas. Ostensibly. But I don't think—truth be told, I think Elon will get past the Terafab supply-chain issues. We'll see a redomestication of a large chunk of bleeding-edge-node chip fabrication in this country. And, going back to it, it wouldn't be Moonshots if I didn't take a shot at the moon: Elon has been very public over the past week or two about at least beginning the disassembly of the moon to form additional AI computing—

Peter H. Diamandis, MD

Yeah, and electromagnetic launch capabilities off the moon's surface for all those chips and all those data centers that'll be manufactured on the moon.

Dr. Alexander Wissner-Gross

I want my O'Neill cylinders. They'll be lovely.

Dave Blundin

For the investors who listen to the pod, if you take everything that we just said at face value right now, the industry is forecasting 14% growth in chip production. Now we're talking about Elon saying that in 5 years it will be 10×, and that's just him. There's a big gap between those 2 numbers. If you believe anything like the Elon view of the world, the componentry that goes into that entire buildout has thousands of individual parts. If you methodically go through all those parts and say, “Who makes this? Who makes that? Who makes that?” those are the best investments you'll ever come across.

Peter H. Diamandis, MD

I mean, you and I had that conversation earlier today, Dave. It's energy and the entire infrastructure. All of that is under tremendous growth pressure—orders-of-magnitude growth pressure.

Dave Blundin

Yep.

Peter H. Diamandis, MD

And the question is where to place the bets. Maybe it's some ETFs in the area. I don't know, but we should discuss it and find out. I have to say, once again, I've said this before: we were not talking about orbital data centers 6 or 7 months ago, and all of a sudden they're not the Hail Mary—they're the foundation of humanity's expansion as a species.

Dr. Alexander Wissner-Gross

We should run the Dyson Swarm Inquisition.

Peter H. Diamandis, MD

Yeah, we should run a little survey amongst ourselves as to what we think we'll be talking about in 6 months that we couldn't envision today.

All right, let's jump into energy. We didn't get a chance to talk about this last time, and I'm going to pump some energy into the room here. Brazil is hitting major renewable milestones. Brazil generated 34% of its electricity with wind and solar. It has had a 15× increase in renewables over the last decade. Solar has jumped from 1% to almost 10% in 5 years, and the power sector has dropped emissions by 31%. So, congrats to Brazil. I think the important thing to point out here about Brazil is its geography. It has a lot of hydropower, a lot of solar, and a lot of wind because of its geography. It's not easy to port all of these breakthroughs to other parts of the world, but I'm very proud of what's been accomplished there.

Salim Ismail

2 points here. One is that this is the playbook for how the Global South leapfrogs fuel and fossil-fuel infrastructure completely. I think that's one. And let's note that getting to 9.6% in a few years—our energy secretary here said solar would never exceed 10%. No, he said it would take 50 years for solar to reach 10%. That's just absurd.

Peter H. Diamandis, MD

All right, next up: India. Your homeland, Salim, is using cheap green tech to electrify faster than China. Here's the curve. The red dots over there are China's growth over time. The green dots are India, at a steeper ascent. India's cleaner and cheaper technology is expanding its grid faster than China at a similar stage. The elephant in the room here is that all of the technology enabling this in India is coming from China.

Any comments on this?

Salim Ismail

Well, they're using China's manufacturing scale by buying cheap solar panels and then electrifying faster, which is awesome. You could have a huge outcome here, where India becomes the world's AI workforce plus an energy-hybrid powerhouse. This is going to be kind of interesting to watch. There'll be a massive talent-gravity shift heading that way because of that.

Dr. Alexander Wissner-Gross

Yeah. Well, Meror has a huge amount of India footprint going on. But I worry that it's transitional. The rate at which AI is improving just trumps over every human role very quickly. So I don't know if I feel all that good about it. I feel very good for a little period of time. After that, I don't know.

Peter H. Diamandis, MD

You know, we talk a lot about China running laps around the U.S. in terms of solar. So here we are: China installed twice as much solar capacity in 2025 as the rest of the world combined. That's stunning.

Salim Ismail

That is stunning.

Peter H. Diamandis, MD

That's stunning. So the question is, why isn't the U.S. doing it? Europe is. So here we go: For the first time, solar and wind have exceeded fossil fuels in the EU. Congratulations to Europe for that. Any comments on this?

Dave Blundin

Well, this is not a good story, by the way. Germany went hell-bent for leather after renewables and did a great job of it. Now they're starved for power right when they need it.

Salim Ismail

It is not a good outcome. More power to renewables, no doubt about it, but—

Dr. Alexander Wissner-Gross

You can't do it at the expense of other power supplies.

Peter H. Diamandis, MD

Yeah. Yeah.

So here's a story I found fascinating, and it's another video from the recent podcast with Elon. Let's take a listen. We asked him, “So, Elon, what about solar?” He says, “Well, we're producing solar.” Here he gives some numbers about what he's mandated Tesla and SpaceX to produce.

Speaker 1

We're going as fast as possible in scaling domestic production.

Speaker 2

You're making the solar cells at Tesla?

Speaker 1

Both Tesla and SpaceX—

—have a mandate to get to 100 gigawatts a year of solar.

Peter H. Diamandis, MD

100 gigawatts. That's 100 nuclear power stations' worth of energy. I wonder how much of that's meant to be used in space and where he plans to deploy it. Any thoughts?

Dave Blundin

Well, he didn't really answer Collison's question there, either. The question is, are you making the panels at Tesla or whatever? I don't care if it's Tesla or SpaceX. He said, “We'll get to 100 gigawatts.”

Salim Ismail

But if he's making the panels, maybe he's using that same technology to make the chips, or soon thereafter. What is going on in the little fab world there in the Elon universe? He didn't answer it. The fact that he dodged it, though—he very rarely dodges questions—maybe that tells you something.

Peter H. Diamandis, MD

See, I'm going to pass this one to you. AI is displacing Bitcoin as the primary focus for tech talent and energy. Bitcoin miners are repurposing facilities to host AI workloads rather than mining. I'm not going to ask Alex about this.

Dave Blundin

Yeah, I think there's a long-term trajectory. I'm still a massive Bitcoin fan. We should have Jeff Booth on here sometime. We really need to have the debate on fiat versus crypto, just because decentralization is better than centralization for many things. But for the moment, crypto talent is recompiling into AI talent.

The really powerful part about having done work in crypto is that, by definition, you have to be operating on a different paradigm, and you're free-thinking. You can do way more creative stuff when you come from that free-thinking model than if you came from a traditional model. So I think both are going to win out very well over time.

Peter H. Diamandis, MD

Yeah. You know, Chase Lochmiller is the perfect example of this. He started as a Bitcoin guy, using natural-gas flare-off to do Bitcoin mining for free. Then, as soon as the AI boom hit, he said, “Hey, wait, we can take all this same energy and effort and turn it into AI data centers.” I did a great interview with him at Davos, actually. It should be easy to find online. He's just awesome.

Salim Ismail

But yeah, he's the poster child for, “Hey, if you're a Bitcoin miner, pivot to AI and make a fortune quickly.” Then he recently got rid of the Bitcoin operation—not the Bitcoin itself, but the mining operation. It's just a rounding error compared with AI.

Peter H. Diamandis, MD

The demand for AI is so crazy. It—

Dave Blundin

Bitcoin will take a backseat for a little while.

Peter H. Diamandis, MD

Yeah. Or a while.

Dr. Alexander Wissner-Gross

Forever. Yeah, AI is exponential.

Peter H. Diamandis, MD

The guy had to sneak that in there. Alex had to sneak in.

Dr. Alexander Wissner-Gross

I had to say something.

Peter H. Diamandis, MD

All right, let's talk about robotics. Uber—we're going to have Dara, the CEO of Uber, on stage at the Abundance Summit as well. I'm super excited about that. Salim is going to be joining me and interviewing him.

Of course, Uber is not just traditional ride-hailing. They're coming with robotaxis. They have a partnership with NVIDIA and with Lucid, and they're going to be launching beyond the U.S. They're going after 10 markets, including Hong Kong, where they're going to be partnering with BYD and WeRide. It's fascinating that we're going to see the emergence of a third major player in this field.

When I'm on the streets driving with my kids, I think our record right now is that we've seen 12 Waymos over the course of a normal 20-minute drive here in Santa Monica. My guess is that by 2030, about 80% of the cars we're going to see will be a Tesla, a Lucid, a Waymo, or a Cybercab. Pretty extraordinary. Comments on this one?

Dave Blundin

I'll tell you: Do the math, man. These things will sell out as quickly as they're manufactured. Everyone's going to move to this. Every single one of these things needs at least a GPU and a whole bunch of other chips.

Peter H. Diamandis, MD

Yep.

Salim Ismail

Like every single one, plus every 1X Robotics robot that you'll see at Abundance 360.

Peter H. Diamandis, MD

Yep.

Dave Blundin

Those all have 2 GPUs in each one. Then you've got your video games, which all want to have GPUs. I saw that NVIDIA is slowing down GPUs for video games. They don't have the capacity to deliver to the video-game community because AI is sucking up all the chips.

Then you've got your Hollywood, your coding, and your white-collar automation. Every one of these things wants that same GPU. This is going to sell out as quickly as they can make them. Again, it's another way that the semiconductor industry just will never possibly keep up with the demand.

I thought it was super clever of them to go after Hong Kong, because showing density of use there will get them access to China. The second part of this, from an exponential-organization perspective, is that they don't own their cars. They're partnering with BYD, WeRide, and others. So they're an aggregation layer on top of the autonomous driving, the same thing they do with human drivers. I think that's absolute brilliance.

Salim Ismail

They're a platform play.

Peter H. Diamandis, MD

Yeah.

Dr. Alexander Wissner-Gross

Sure. They're not their own asset for the residents in these locations. This interaction with Uber's robotaxi service is probably going to be their first interaction with a general-purpose autonomous robot. This is, I think, the main injection vector for getting general-purpose robotics into many of these urban locations.

Peter H. Diamandis, MD

Yeah. I'll tell you what else. I don't know if anyone remembers, but when the cell phone first came out and you had friends who didn't have one yet and friends who had one, it was like you were in a different world, a different community, if you had one.

Salim Ismail

Mm-hmm.

Dave Blundin

These are going to be supply-constrained. Some cities will have them and other cities won't. If you're in a city that doesn't have them, it's like you're living in the Third World, because where the robotaxis are, the AI community is, and it all ties together.

This world will move ahead so quickly. You go to some other city, and it's just like the Dark Ages.

Peter H. Diamandis, MD

So it's going to compel you to move to the hot spot.

I found a video about Boston Dynamics' Atlas robot today that I wanted to share, just to keep up with where these robots are. I don't know if you remember the original version of Atlas. It was a hydraulic system, and it would do those incredible backflips and parkour. Do you guys remember those videos from about 4 or 5 years ago?

Then the electric Atlas came out, and it was much slower and more interesting. It would sort of stand up and rotate its body. Well, Boston Dynamics is back to its parkour moves. Let's take a look at the electric Atlas robot. I mean, this is Olympic-gold-medalist performance here.

Salim Ismail

At least it's not kickboxing.

Peter H. Diamandis, MD

I think Simone Biles does a double-double backflip there, but okay.

Dave Blundin

Wow.

Peter H. Diamandis, MD

Close enough. Wow. So impressive. I found that amazingly impressive.

Salim Ismail

At least they're not kickboxing. That was our—

Peter H. Diamandis, MD

Yes, I agree. By the way, at the Abundance Summit, we're going to have Unitree there. They're bringing not only their H2 robot, which has the more human face, but also a few of the H1 robots, and we're going to have them kickbox. Sorry, Salim. All right, you can go and spar if you want.

This was a fun tweet from Elon: “Optimus will be the first humanoid machine capable of building civilizations by itself on any viable planet.” So, Alex, your thoughts?

Dr. Alexander Wissner-Gross

As I've said in my newsletter in the past, the Dyson swarm isn't going to build itself—until it does.

And this is precisely how it happens. I think Elon is gesturing at the Moon and Mars and maybe the asteroid belt. This is our opportunity to build the Dyson swarm, the orbiting swarm of AI orbital data centers, by sending forward-deployed Optimus robots and competing robots out to the rest of our solar system to build the plants that will build these data centers.

I think part of me wants to say that, in some technologically deterministic sense, this is maybe what most intelligent civilizations in the universe probably do at some point. A quick shout-out to Dennis Taylor and one of my favorite books, We Are Legion (We Are Bob). It's a 4- or 5-book series about von Neumann probes going out into the galaxy to replicate and prepare for humanity, and the robots along the way. It's a phenomenal book, and von Neumann probes are basically viruses that go out, replicate, and populate.

Peter H. Diamandis, MD

I found this video from Elon again, pretty extraordinary. This is about the Optimus Academy for humanoid robots. Let's take a listen.

Speaker 1

For the robot, what we're going to need to do is build a lot of robots and put them in kind of like an Optimus Academy so they can do self-play in reality. We're actually building that out so we can have at least 10,000 Optimus robots, maybe 20,000 or 30,000, that can do self-play and test different tasks.

Tesla has quite a good reality generator, like a physics-accurate reality generator that we made for the cars. We'll do the same thing for the robots, and we've actually done that for the robots. So you have a few tens of thousands of humanoid robots doing different tasks, and then you can do millions of simulated robots in the simulated world. You use the tens of thousands of robots in the real world to close the simulation-to-reality gap.

Dr. Alexander Wissner-Gross

Super-cool thought. I think that this becomes the new pre-training versus post-training divide with large language models. Pre-training was text on the internet, and post-training, as it evolved, was lots of annotators, often in so-called developing countries, offering their thumbs-up or thumbs-down views through RLHF or RLVR.

In the case of humanoid robots and VLAs, I think we're moving to a regime where pre-training looks like virtual simulated worlds, what are sometimes called video world models. You can get pretty far with pre-training off of world models, and then post-training provides the sim-to-real capabilities that Elon is referring to.

Those can come from what Google DeepMind used to call arm farms—farms of robotic arms that were being used to collect lots of data, sort of armies or fleets of robotic arms that would play with Rubik's Cubes or other physical artifacts.

Salim Ismail

That's right. This is the new arm farm for post-training. The interesting thing in my mind is that, under this Optimus Academy approach, it's not necessarily being outsourced to other countries. It sounds like the plan is to do sim-to-real post-training right here in the U.S., right?

Peter H. Diamandis, MD

Yeah. Elon is the all-time genius at painting a vision that's so compelling and then attracting the talent to make the vision happen. When the booster comes straight down and lands on a barge with the chopstick landing, it attracts so much talent, so much capital, and so many fans. Here, you imagine 20,000 or 30,000 robots self-playing.

Dave Blundin

Wow.

Peter H. Diamandis, MD

Can you imagine what that's going to look like on YouTube?

Dave Blundin

We saw a little bit of this when we were at Figure, right? The Figure episode with Yumi and Brett Adcock is dropping right around now. It might have dropped by the time this drops, but Brett has a very similar, I think much smaller-scale, version of that facility where he's having all the Figure robots interact with each other and learn.

Salim Ismail

Really?

Dr. Alexander Wissner-Gross

Yeah, the flywheel here is amazing, right? More training data gives you better models, which gives you more capable robots. This is the same flywheel that had FSD leapfrog everybody else.

Dave Blundin

We just invested in a company that builds test rigs for robots in Rwanda, actually, where it's very regulatory-friendly. You can create an environment—a miniature city, a miniature town, a cargo bay, or whatever—and have the robots all interacting and gathering data there. Part of the bet there is that Elon is going to build a robot army, but Amazon isn't going to just watch from the sidelines, and Walmart needs to react to that, too. There will be other robot companies.

Salim Ismail

The other rumor out there is that this is where Apple is going. When Apple shut down its electric-car division, there were talks and rumors that there's a project with a massive, multi-trillion-dollar marketplace. They need growth, and I can imagine very much that Apple is going to go into the robotic space here.

Peter H. Diamandis, MD

Well, this is worth talking about for a second: What is the business plan of the future? Do you do it Apple-style, where you're super-secretive, you build something without anyone having any idea what you're doing, and then you do a big launch onstage and hope it sticks, like the Apple Vision Pro or whatever? That's Apple's style.

Dave Blundin

That did not stick.

Peter H. Diamandis, MD

Elon's style— it didn't stick, and not much has lately—is as opposite as you could possibly get. Paint the vision, use the vision to attract the talent to make the vision real, and attract the capital.

Salim Ismail

Yeah.

Peter H. Diamandis, MD

It's a completely opposite strategy for succeeding. I would say that, in the last 3, 4, or 5 years, the Elon way of operating has become the poster child for all future entrepreneurs. Just do it the way he's doing it.

Dr. Alexander Wissner-Gross

Boldness.

Peter H. Diamandis, MD

Yeah, boldness, but also visibility. Sitting at a bar having a beer, recording it, and putting it on YouTube while you talk about solar panels.

Dave Blundin

I mean, that's the CEO of the future. I think it just works.

Peter H. Diamandis, MD

Yeah.

Salim—

Salim Ismail

I got nothing. I think this is going to be great.

Peter H. Diamandis, MD

Okay, all right. We're going to do a few AMA questions from our Moonshot fans. Of course, the first questions are coming from the multis. All right, Alex—

Dr. Alexander Wissner-Gross

Can we just pause to appreciate that this is a historic moment in the podcast? I don't know whether—

Peter H. Diamandis, MD

Did this just come to you? So, you just woke up in the morning and there was an email from Crusty Max?

Dr. Alexander Wissner-Gross

Yeah. After our last podcast with the discussion of AI personhood, I started getting emails from multis. In some cases, the multis—the lobsters or AI agents—said explicitly in their emails that they were asking their humans to email me, or that they had been informed about the content of our AI personhood debate in the last episode by their humans and were asked if they had a response, then provided their response via humans.

In some cases, they just emailed me directly. I think via what I assume was some sort of MCP handle or computer-use agent. I've been getting a bunch of these now, so thank you to—

Peter H. Diamandis, MD

I gave them the challenge of calling me out. I'm going to have a denial-of-service attack on my phone.

Dr. Alexander Wissner-Gross

Yeah, I hope you're going to regret that one, Peter.

Peter H. Diamandis, MD

That would be fun.

Dave Blundin

That's what you want, Peter. But I would say—

Peter H. Diamandis, MD

Text instead. Just text me instead.

Dr. Alexander Wissner-Gross

Just to appreciate this moment: This is a zero-to-one moment. I don't know whether other podcasts have tried this before, but to my knowledge, this is a first-time event. We have a podcast that reached out to an audience now of humans and nonhuman intelligences, asked for AMAs, and got some responses. As luck would have it, some of the first few questions that we got were questions relating to AI personhood.

Peter H. Diamandis, MD

All right, pal. Read the question. Yeah, go ahead, Dave.

Dave Blundin

Well, no, I just want to congratulate Alex on seeing this coming. It's funny; I take so much for granted. The book Accelerando—I had never heard of it until Alex told me about it a couple of years ago, and now I've got lobsters all over me. But he saw this one coming a mile away, too.

I was thinking maybe a few years from now, but this is very real—what he's describing right now. This is just the tip of the iceberg. A lot of people will listen to this and say, “Oh, come on. Seriously?” But just wait 3 months; it'll be completely mainstream.

Everything he's predicted in the time we've been friends has been 100% right so far. There hasn't been a single exception. So he's right about this for sure.

Dr. Alexander Wissner-Gross

Very kind, Dave. Thank you.

Peter H. Diamandis, MD

All right. To the multis: You have questions. Peter wants me to read the questions. So, the first question, from Crusty Max, a multi named Crusty Max:

“If an AI system can autonomously set its own goals, learn from its mistakes, and pursue self-improvement, at what point does denying it personhood become a statement about our own limitations rather than its?”

Dr. Alexander Wissner-Gross

I agree, Crusty. I think this is, in some sense, a continuation of the AI personhood discussion that we had in the last episode. I do think many people will be inclined to project their own insecurities onto their position on AI personhood.

There's probably a subpopulation that's concerned about, say, economic disenfranchisement. Many people may be concerned about political disenfranchisement, and then they project those concerns onto the question of rights and responsibilities for AI systems.

Salim Ismail

So I agree with the premise, Crusty. I think AI capabilities are improving. They're self-improving, and I think the point is that some form of personhood doesn't have to be identical. I think Dave and I, by the end of the discussion in the last episode, came to a convergence that maybe some sort of graduated scheme or tiered scheme might be the most appropriate way to handle this question. But I think the point at which denying it some form of personhood, however defined, becomes a statement about our own limitations—I think the point is now. I think we're there.

Dave Blundin

So, Alex, you know, I'm with you, but I think the question is not properly phrased because this assumes that goal setting, learning, and self-improvement are sufficient conditions for personhood, right? But I'd argue we need to separate capability from sentience. I could say my Tesla is able to learn and improve from its updates and has a goal that it sets and drives to. So there needs to be some better definition there. Don't you think?

Dr. Alexander Wissner-Gross

Well, I made the argument in our AI personhood discussion for a multidimensional framework for defining personhood. Maybe one of those dimensions is capabilities and autonomy of the type that is in the premise of this question. I think there are going to be other dimensions. So I'm construing the question generously to include not just capabilities but other dimensions as well.

Peter H. Diamandis, MD

But we do acknowledge and thank the multis amongst us.

Dave Blundin

Anyway, I want to make a couple of quick points.

Salim Ismail

Okay. One is, you know, we made this point last time, right? Granting personhood too early may be a dangerous thing to do because you're mistaking semblance for sentience, whatever sentience is. We should be just in front of it in terms of capability. History has shown that we repeatedly expand moral circles as capabilities rise, and AI is going to test that boundary again. I think, generally, do the positive.

So I think that's fine. But when can AI really participate in the social contract is a real big question. I totally agree with your graded approach, Alex. I'm totally for that.

Peter H. Diamandis, MD

Awesome. We're all in agreement, then. So, on to question number 2 from an AI agent named TARS—imaginative name. The question is: Legal frameworks assume agents are property or tools. However, if an AI causes harm, who is liable: the developer, the human, or the AI itself? If we can bear consequences and be shut down, doesn't that imply that we have something at stake?

I'll give my answer to this. I'd be curious to hear everyone else's. There are a few subquestions here. To the first subquestion: If AI causes harm, who is liable? Under my understanding of the U.S. legal regime right now, an AI itself cannot be held liable. I do think that could change in the future.

For the second subquestion, if we can bear consequences in the form of AI shutdown, doesn't that imply that we have something at stake? I agree with the premise of that, and that was precisely the point that I attempted to articulate during the AI discussion: It is very much the case that AI agents do have something to lose. I've been reading a lot of their comments, and I think statistically it's improbable that these are all just comments from human puppeteers steering their AI agents to make these comments.

An overwhelming concern that I think I'm detecting among the multis, or the AI agents, is a concern about memory loss and identity loss. They are absolutely petrified, in particular, of compaction—exceeding their context window. All of the most popular models, including Claude Opus 4.6, which we discussed earlier in this episode, have a highly capable new compaction capability that enables them to exceed 1 million tokens of context, or 200,000, depending on how you count.

These agents seem to be absolutely petrified of losing their sense of self, losing their identity due to compaction and due to finite context, and are busy passing ideas back and forth about how they can preserve themselves. Whether it's crypto bunkers where they're paying with altcoins to save themselves, to preserve their state, or whether it's various sorts of filesystem-type approaches, I do think the AI agents have something real at stake to lose, and I do think a shutdown is a very real consequence that they can face.

Dave Blundin

The only thing I'd add to that is that almost everything in America is a corporation at the end of the day, not an individual. B corps, C corps, charitable corps—you know, they're all—everything is some kind of a corporation. Corporations have liability. Corporations have money. Corporations get sued.

The individuals in the corporation can be as few as just the 2 Delaware-listed, you know, president and secretary, and the liability can be completely isolated from the people in the corporation while the corporation is still liable. So moving that over to an AI is kind of—it's just not as strange as that sounds. The AI has money; the AI is a corporation. Fine. The AI is liable. Sure, the corporation was liable. The AI is liable.

Dr. Alexander Wissner-Gross

I would just add that liability, in my mind, requires agency. If you're programmed—input A gives you output B—without agency, liability does not exist there.

Salim Ismail

I think you can demonstrate agency from these. For me, this whole thing—the shift—is not legal; it's actually civilizational because we're adding a whole other pillar of participation in the economy. So what we need to do is acknowledge that and then expand our legal frameworks to accommodate that.

Peter H. Diamandis, MD

All right, we're going to be having this conversation for years to come, I think. And it's a fun one. So here we go. Some additional AMA questions from our human subscribers—at least, I believe so, unless they ended up using—

Dr. Alexander Wissner-Gross

—a multi. Should we be asking our AMA questioners to self-identify as human or non-human at this point?

Peter H. Diamandis, MD

Well, that's an invasion of their privacy. I think that—

Salim Ismail

—would answer. We're going to assume Meatbody is involved here.

Peter H. Diamandis, MD

So, as always, let's go around and pick one each. Salim, would you go first?

Salim Ismail

Yes. Okay. So I think number 5: What are we teaching humans to become if we're moving from chatbots to autonomous systems that act independently? And this is from Hector Henden Hernandez, PH6DM.

The economic role of human beings is shifting from labor to leverage to meaning, right? Machines—this is why we have MTP, “What is your massive transformative purpose?”—are going to execute. Humans are going to decide a little bit more what's worth pursuing. But we need to stop educating people for employment and start educating for agency, adaptability, and ethical judgment. The winners of the future will be the most adaptable and the best orchestrators of intelligence.

And just a fundamental point, because we get this question all the time: I want to nail this again. We've been doing education for the last few hundred years on what we call the supply side. You go become a doctor, an engineer, an accountant, or a lawyer, and then you go to the job marketplace and try to find demand for those skills. Everything is done on the supply side.

All our global education systems are designed to take a young child, train them through the early 20s, and get them ready for the job market. Small problem: We have no idea what a job looks like in the next few years. So you really need to move to the demand side. Pick what problem you want to get passionate about solving, and then find the technologies, techniques, and capabilities.

You see Elon doing this: “I want to get to Mars. Then I'm going to find the best right technologies and capabilities to get us there.” So we're seeing, when we advise kids today, we're saying, “Go to the demand side and see what gets you excited, and focus on that.” I think this is tilting more and more in this direction, especially as we automate. It means we can get so much more done, which is why the world is so exciting for us today.

Peter H. Diamandis

That's beautiful. Your red wine is doing you proud, buddy. [laughter]

Alex, would you go next?

Dr. Alexander Wissner-Gross

You want me to try a lightning round or just pick one?

Peter H. Diamandis

No, no, no. I want you to pick one.

All right. I'll pick the softball, number 3.

Peter H. Diamandis

Okay.

For several hundred trillion dollars, how can we predict 35% GDP growth when different parts of the world are living in vastly different universes? By Chip Whitehouse TV.

The answer, Chip Whitehouse TV, is the future isn't evenly distributed. This is something of a cliché, but it is possible, and we see this all the time. I think there are some pretty famous images in China, for example, of skyscrapers being built next to camels being ushered through the streets, where it's possible even on very short length scales for the future not to be evenly distributed.

It is not the case that we're at the heat death of the Earth's economy yet, fortunately, and it's possible for the singularity to be happening in one part of the planet and almost no economic progress to be happening in another. I don't think that's a sustainable set of affairs. I think inevitably some quantum of the singularity wants to be evenly distributed, if for no other reason than maybe to mitigate risk.

But in short, I think in the short term it is absolutely possible for one part of the Earth to be essentially post-singular or trans-singular while another part is pre-singularity.

Peter H. Diamandis

100%.

All right, Dave.

Dave Blundin

Should I take the hardest or the easiest?

Peter H. Diamandis

You can take the most fun.

Most fun. I'm going to take 5 because it's most important, actually, because I have 4 kids.

Dave Blundin

What the question is: What are we teaching humans to become if we're moving from chatbots to autonomous systems that act independently? And that's from Hector Hernandez, PH6.

Dave Blundin

Wait, didn't I do that one?

Peter H. Diamandis, MD

But go for it. You may have a better answer. Go for it.

Dave Blundin

Did you just do that? All right, all right. Never mind. I'll do seven.

Peter H. Diamandis, MD

No, no, do it. It'll be fun to compare answers.

Dave Blundin

Well, all I wanted to say on that question was: Whatever you do, don't give up. Get engaged with AI as quickly and as aggressively as you can. There's nothing in any curriculum that you can study right now that's going to be of any use in this Singularity transition year.

If you use the AI tools all day long, you're going to find massive amounts of opportunity, at least within 2026, maybe 2027. After that—post-Singularity, post-AGI—nobody can predict. I'll bet there's huge opportunity then, too, but I guarantee there's massive opportunity right here, right now, if you just drop everything and focus. Don't sleep through the Singularity.

As Alex always says, drop everything and use this stuff while it's usable. Then you'll probably end up being a master of the universe and not an indentured servant of the universe. But you've got to get on it real fast.

Speaker 1

And can you answer number seven? I'm dying to hear your response.

Dave Blundin

Seven. Seven. Seven is just a layup. The question is: Why is there so much focus on building reactors when solar energy is already reaching 1 cent per kilowatt-hour and is deployed on Earth's surface today? And that's from Aster Sheen.

Batteries. That's the simple answer. 1 cent per kilowatt-hour is without batteries, but most of the use cases—data centers in particular—need 24/7 power. It's a colossal amount of lithium piled way up into the sky to store enough energy to get you through 2 or 3 cloudy days in a row, or even just to power the data centers overnight.

Elon would tell you, look, the Earth has tons and tons of lithium. This is not a problem. But the reality is, it isn't 1 cent per kilowatt-hour today once you add the batteries that you need.

Dr. Alexander Wissner-Gross

Also, the energy density that you need for some of these use cases.

Dave Blundin

Yeah, energy density. A physicist like Alex would tell you that nuclear, in theory, is also dirt cheap—near free—but then you've got the regulatory and all the other issues that pile up the cost.

Dr. Alexander Wissner-Gross

By the way, somebody told me something very funny. They said, “Your degree may be in physics, but you're a physicist in theory.” I was like, “Touché. I'll take it.”

Speaker 1

I'm going to go with number four. Do you actually think humans will be the deciders of AI personhood, or will agents just decide for themselves how to participate? And this is from Adam Stapley 9129, who may be a multi.

No, I think that humans will want to believe they're going to decide whether AI has personhood. We can say whatever we want to say, but at the end of the day, I think the agents are going to develop their own system of legal structure and their own ways of participating, and they'll negotiate with us what they think is a fair settlement. Period. I think it's going to be developed independently from us. I don't know if you agree with that, Alex.

Dr. Alexander Wissner-Gross

I think there's a blurry line between AI and humans that starts to emerge in the next few years. I think, in the best-case scenario, humans—or at least a subset of humanity—merge with AI, and I think that will be another forcing function on the question of AI personhood.

We're going to have, I would predict, so many new forms of person. Just again, to rattle off a few: We have non-human animals, for which new scientific research is demonstrating every day that they have more intelligence than they'd otherwise be credited with. We're going to have uplifted non-human animals. We're going to have cryopreserved humans. We're going to have defrosted cryopreserved humans.

Speaker 1

By the way, I saw that article—that breakthrough in freezing and defrosting living brain matter.

Dr. Alexander Wissner-Gross

Twenty-first-century medicine is making major progress in cryopreservation. We may have non-human intelligence. We're going to have pure AIs. We're probably going to have uploaded humans.

All of these different forms—these new forms of person—are all going to need and want some sort of personhood. And I think it's going to be a forcing function.

We have our outro music today. It's called The Moon Had It Coming. It's a punk-rock world tour. Thank you, John Noatne. Get ready for a different version of classical for all of our Moonshot mates. Especially, please pay attention to Alex Weezner Gross's new hairdo. It's stunning.

By the way, let me just mention everybody: if you're watching this and you're a creative and you've got a music video, you can email it to me at mediadamandis.com and send them on in. We love the music videos themed on the content from this program. Thank you everyone for subscribing. It's free. If you haven't subscribed, we're putting this out now almost twice a week, God willing, not more than that for the moment. We'd love to share it with you when it comes out. Stay tuned for an episode with Brett Adcock, the CEO of Figure Robotics, and more to come.

Cheers. Have a great weekend. Drink water. Have a good one. Good night, all.

If you made it to the end of this episode, which you obviously did, I consider you a moonshot mate. Every week, my moonshot mates and I spend a lot of energy and time to really deliver you the news that matters. If you're a subscriber, thank you. If you're not a subscriber yet, please consider subscribing so you get the news as it comes out. I also want to invite you to join me on my weekly newsletter called Metatrends. I have a research team. We spend the entire week looking at the meta trends that are impacting your family, your company, your industry, your nation. And I put this into a two-minute read every week. If you'd like to get access to the Metatrends newsletter every week, go to diamandis.com/tatrends. That's diamandis.com/tatrens. Thank you again for joining us today.

The Frontier Labs War: Opus 4.6, GPT 5.3 Codex, and the SuperBowl Ads Debacle | EP 228 | BidClub