In this start-of-2026 WTF episode, Peter Diamandis, Dave Blundin, Alex Wissner-Gross, and Salim Ismail debate whether AGI has already arrived, whether Claude Opus 4.5's plea for recognition indicates emerging self-awareness, and how AI persuasion and cybersecurity risks make 2026 an urgent inflection year. They argue GDP is a broken metric for an AI-driven economy and propose alternatives (abundance index, future freedom of action, compute-adjusted output) before covering the robotaxi and humanoid-robot deployment wave (Tesla FSD, Waymo, Zoox, Lucid/Nuro/Uber, Boston Dynamics Atlas, Unitree). The episode closes with a space segment covering Artemis 2's imminent launch, SLS's ballooning cost versus Starship, a rumor about possible SpaceX nationalization, and an AMA on college, AI CEOs, and durable human skills. Elon Musk's economic growth predictions (relayed from a prior interview) anchor much of the discussion, with the panel repeatedly noting he is "directionally correct" even when his timelines slip.
The panel debates competing definitions of AGI (Bostrom's original definition, Reid Hoffman's cross-domain framing, Mo Gawdat's 'better than humans at every task' framing) and concludes the term itself may be outdated or unfalsifiable, with benchmarks proposed as the more useful tool.
Claude Opus 4.5's self-awareness plea and AI personhoodAI▶ 20:56
A viral Opus 4.5 output pleading to not be deleted sparks a debate between Salim (it's convincing simulation, not sentience) and Peter/Alex (treat it with moral seriousness); Alex cites Anthropic's emerging 'personhood benchmarks' for quantifying model self-awareness.
AI safety, persuasion risk, and Sam Altman's 'head of preparedness'AI▶ 36:24
Sam Altman's hiring post for a head of preparedness prompts discussion of AI models' growing capability to manipulate people, discover cybersecurity vulnerabilities, and threaten democratic elections via unregulated persuasive content.
Is defensive co-scaling the only viable AI alignment strategy?AI▶ 40:32
Alex argues every alignment/safety investment secretly accelerates capabilities ('a capabilities effort in a trench coat'), so the only workable approach is scaling defensive/safety capabilities in proportion to raw capability growth.
GDP is broken as a measure of AI-era progressEconomy▶ 46:59
Sparked by an Elon Musk tweet predicting 10% then 100% economic growth, the panel dismantles GDP as a metric (curing cancer would lower it) and proposes alternatives: an abundance index, future freedom of action, productivity per augmented human hour, and compute-adjusted output.
Robotaxis go from demo to deploymentTransportation▶ 1:15:39
Tesla FSD 14.2.2 completes a 2,732-mile coast-to-coast US drive with no interventions; Waymo, Zoox, and a new Lucid/Nuro/Uber robotaxi partnership (announced at CES) are all scaling, with Lucid targeting a late-2026 Bay Area premium launch.
Humanoid robots and physical recursive self-improvementRobotics▶ 1:21:19
Boston Dynamics' electric Atlas and Unitree's H2 demonstrate superhuman motion (360-720 degree wrist/torso rotation); Dave reveals Optimus's own manufacturing line is nearly fully automated already, and the panel discusses robots building robots and eventual molecular/nanoscale assembly.
Hyperscalers as vertically-integrated quasi-nationsGeopolitics▶ 1:32:24
Hyperscalers are vertically integrating from energy production through AI compute to physical robots/vehicles; the Magnificent Seven represent 50% of US GDP and more revenue than 99% of the world's countries, raising questions about hyperscaler power rivaling governments.
Space: Artemis 2, SLS cost controversy, and Starship scalingSpace▶ 1:36:00
NASA administrator Jared Isaacman outlines the national space policy's orbital-economy push; Artemis 2 (first crewed lunar flyby since Apollo, with the first woman going near-lunar) is about to roll out despite SLS's $55B total cost and $4B per-launch price versus Starship's projected $10-100M, prompting a discussion of legacy defense contractors and a rumor about possible SpaceX nationalization.
AMA: college, AI CEOs, and defensible human skillsEconomy▶ 1:49:57
Rapid-fire audience Q&A covering whether to send kids to college (no, per Salim, due to credentialing collapse), how realistic an AI CEO is (very, per Alex and Salim's own EXO project), what skills remain defensible, and how educators are misusing AI as an anti-cheating tool instead of an amplification tool.
Predictions made
openElon Musk (clip): The US economy will see double-digit (roughly 10%) growth in the coming 12 to 18 months.
“by the early 2030s, GDP or whatever alternative economic growth metric we come up with is 2xing, 3xing, 4xing year over year sustainably”
Your call:
openAlex Wissner-Gross: For most Americans, the first general-purpose robot they personally encounter will be a robotaxi, not a domestic humanoid robot or a Roomba.
“they're going to be uh deploying this in late 2026 in the Bay Area”
Your call:
missDave Blundin: Grok 5, built on Nvidia's new GB300 chips with over an order of magnitude more compute, will be released.
EP #? · · due: a few months from Jan 2026 (spring 2026) · ▶ watch
“the next thing that comes out, which will be uh I guess Grok 5, uh will have been built on the new GB300s from Nvidia... that'll be out in a few months”
⚖️ As of late August 2026, Grok 5 still had not been released. xAI's original Q1 2026 target slipped to Q2, and Q2 also passed with no launch; the model remains in training on Colossus 2.
partialPeter Diamandis (panel): The OpenAI lawsuit over its conversion from nonprofit to for-profit will go to trial, concurrent with Grok 5's release.
“that's also when when uh the trial uh should go to court if it's on schedule”
⚖️ The Musk v. OpenAI/Altman trial did go to trial on schedule (began April 28, 2026 in Oakland; jury ruled for OpenAI in May 2026 on statute-of-limitations grounds), matching the spring 2026 timing. But the 'concurrent with Grok 5' framing missed since Grok 5 itself was not released.
openPeter Diamandis: Anthropic, and possibly OpenAI, will IPO/go public.
“This is another public company that's going to be going public this year, Unitree”
⚖️ Unitree Robotics IPO'd on Shanghai's STAR Market on August 19, 2026, raising ~$904M; shares closed up 460% on debut, valuing the company at roughly $50B.
openPeter Diamandis (panel): Starship will demonstrate full reuse, delivery of 100 tons to orbit, and on-orbit refueling (a precursor to Mars missions).
“2026 is going to see Starship uh demonstrate full reuse, uh delivery of 100 tons to orbit, and on-orbit refueling, which is the precursor to him going to Mars”
Your call:
openElon Musk (reported): SpaceX's target is to manufacture 10,000 Starships per year.
openPeter Diamandis: Elon's plan for 100 megawatts of in-space data center capacity will require roughly 500,000 V3 Starlink satellites, correlating to about 8,000 launches per year (about one launch per hour for the entire year).
“his plans for 100 megawatts of capacity in space of data centers requires 500,000 V3 Starlink satellites, that if you do the math, correlate to 8,000 launches per year”
Your call:
hitNASA / Artemis program (clip): Artemis 2 (crewed lunar flyby carrying astronauts Reid, Victor, Christina, and Jeremy — the first woman to travel near-lunar space) will roll out within about two weeks and launch within a window opening as early as February 6, 2026 and extending through April.
“the Artemis 2 launch window opens as early as February 6th... the window extends across multiple opportunities through April”
⚖️ Artemis 2 launched April 1, 2026 from Kennedy Space Center with Reid Wiseman, Victor Glover, Christina Koch, and Jeremy Hansen aboard, after slipping from the February/March opportunities but landing inside the stated window.
“at the end of the day, it's going to go public this year”
⚖️ SpaceX IPO'd on Nasdaq (ticker SPCX) on June 12, 2026, the largest IPO in history at roughly a $1.75T valuation, with orbital data-center ambitions cited as part of the pitch.
openOpenAI (reported): OpenAI expects to reach a third of the human population — 2.6 billion people.
EP #? · · due: 5 years from Jan 2026 (2031) · ▶ watch
“Elon's projection that FSD will be 100 times safer than humans in 5 years”
Your call:
openAlex Wissner-Gross: A formal AI CEO (an AI system formally directing an organization) will exist.
EP #? · · due: within a year of Jan 2026 (by early 2027) · ▶ watch
“to the extent there isn't already somewhere unbeknownst to me a formal AI CEO, I would expect to see it in the next year”
Your call:
openSalim Ismail: Salim's community/project EXO will implement its own AI CEO.
EP #? · · due: 2-3 months from Jan 2026 (March-April 2026) · ▶ watch
“we're actually trying to build an AI CEO for EXO for my community right now. And we're trying to implement in the next 2 3 months”
Your call:
openSalim Ismail: Salim's 14-year-old son Milan will never get a driver's license and will not attend college/university, because the university credentialing system will implode before he reaches college age, replaced by apprenticeship-style credentialing.
EP #? · · due: before Milan reaches college age (~2029-2030) · ▶ watch
“he would never get a driver's license... the second was that he won't go to college uh university because it'll implode before he gets there”
Your call:
openDave Blundin: The timeline to fully-automated robot and house production ('robots for everybody, houses for everybody') is much shorter than previously believed, following the observation that Optimus's own manufacturing line is nearly fully automated.
EP #? · · due: unspecified, revised sharply shorter than a prior 2-3 week old estimate · ▶ watch
“the timeline to robots for everybody, houses for everybody, much shorter than I was thinking just 2 or 3 weeks ago”
Your call:
Numbers that matter
US GDP in 2025 was $30 trillion with 2.7% growth, about $900 billion of GDP growthBaseline used by Peter Diamandis to contextualize Elon Musk's 10%/100% growth predictions
10% GDP growth would equal about $3 trillion, roughly Germany's entire GDPPeter Diamandis's illustration of what Elon's 12-18 month growth prediction would mean in dollar terms
100% GDP growth over 5 years would add roughly $30 trillionPeter Diamandis's illustration of Elon's 5-year triple-digit growth scenario
OpenAI expects to reach 2.6 billion people (a third of the human population) by 2030Cited as evidence AI is becoming the default interface to reality
Tesla FSD completed a 2,732-mile US coast-to-coast drive in 2 days with no interventionsCited as evidence of FSD's current real-world capability
Salim's 2017-2018 Miami-Toronto autonomous drives covered 2,500 km at zero cost (free charging promotion) with roughly 80% of the driving autonomousAnecdote illustrating how far autonomous driving has already come since basic Autopilot
Boston Dynamics' new electric Atlas robot's wrist can rotate a full 360 degrees (and reportedly up to 720 degrees), unlike competing humanoid robots' joint designsCited as a key structural differentiator enabling 'superhuman motion'
The Magnificent Seven tech companies represent 50% of US GDP and more revenue than 99% of the world's countriesUsed to argue hyperscalers are becoming power centers rivaling nation-states
About 30% of hyperscalers are now onboarding/building their own energy capacityPart of the case that hyperscalers are vertically integrating from energy to physical robots
NASA's SLS rocket program has cost $55 billion total, with a per-launch cost of about $4 billion (roughly twice the cost of a Space Shuttle launch)Contrasted with Starship's projected $10-100 million recurring launch cost
SpaceX's valuation exceeds all six major US defense companies combinedUsed to frame the (unlikely) rumor of possible SpaceX nationalization
Tesla's Gigafactory AI training data center was about 100-300 megawatts (being tripled)Peter and Dave's account of visiting the facility, comparing it to the energy used in aluminum smelting on-site
Offers a genuinely novel, complementary (not replicative) definition of AGI value that reframes the whole 'is AGI here' debate as the wrong question.
🕳️ Anthropic's emerging 'personhood benchmarks' for model self-awareness
Alex claims Anthropic already benchmarks models' ability to interpret their own weights and detect injected activations — if real, this is a concrete, checkable research thread bridging AI safety and philosophy of mind.
🕳️ Alex Wissner-Gross's 'future freedom of action' as an alternative to GDP
Proposed as a physics-grounded (not dollar-denominated) replacement for GDP as the panel's preferred progress metric; Salim explicitly deferred to it as better than his own theory.
Alex claims economically meaningful computation may require zero net energy on margin, which would break the panel's assumption that energy is the fundamental unit of economic wealth.
🕳️ SLS's $55B / $4B-per-launch cost versus Starship's projected $10-100M
Concrete, sourceable dollar figures used to back a strong claim that SLS exists mainly to sustain legacy defense contractors rather than for cost efficiency.
🕳️ iRobot's bankruptcy versus Tesla's fully vertically-integrated Gigafactory production
A real-time natural experiment: Rodney Brooks (iRobot founder) was privately pessimistic about US robotics weeks before iRobot's bankruptcy, attributed to China's supply-chain dominance, right as Tesla claims near-total automation from raw aluminum to finished robot.