Updated 2026-11-16 (weekly check: on EP310 and EP311, both already read below, now backfilled to §11; Stealth 42 Ltd has no filing beyond its incorporation; II blog, Champion terms and thesis unchanged; no open call). Earlier: 2026-11-10 (★ Stealth 42 Ltd incorporated, him sole director and 75%+ controller; not on EP308 or EP309); 2026-11-04 (EP307 read; EP303's video-price call provisionally scored not-met); 2026-10-29 (EP306 read; not on EP305); 2026-10-23 (Hodes opinion read); 2026-10-21 (EP303 read in full, section below); 2026-10-17 (Stability PSC register; Navier–Stokes relay barred); 2026-10-11 (Evoke post, Champions date resolved); 2026-10-05; 2026-10-02, EP294 read end to end (the liability-waiver position, §04 row 12). Refreshed weekly (see How this page stays current at the bottom).
What changed this week (checked 2026-11-16, window 11–16 Nov):
- Moonshots: on EP310 and EP311. EP310 (11 Nov, recorded 10 Nov) and EP311 (13 Nov, live) are both his. Each has its own dated section below, and this week both get their missing §11 rows. Caption-spelling hits in EP311:
Immad×3,Emod/EMOD×2,Mustach×1. Neither tape names the Champion, The Last Economy, ii.inc, the Intelligent Internet or Foundation Coin (grep of both transcripts). The watcher last ran 15 Nov 07:00 and reported "no panel episode aired that we don't hold". - A stealth robotics company, on tape (EP310, recorded 10 Nov). "we even set up a robotics company that I'll tell you guys about. It's a bit stealth at the moment. Like it's going to go crazy." (
1:27:39; no speaker labels, attribution inferred from the hand-off). He gives no name, jurisdiction or date. Stealth 42 Ltd was incorporated the same day (below), but its filing is IT consultancy by SIC code and does not say what the company does, so nothing on the record links the two. Treat them as separate until a filing or a statement links them. - Filings: no change since 10 Nov. STEALTH 42 LTD (17574002) is still active under the same name. Its one filing is still the 10 Nov incorporation (GBP 1, model articles), and he is still sole director and sole PSC (75%+ shares and votes, right to appoint directors). First accounts are due 10 Aug 2028 and the first confirmation statement by 23 Nov 2027. Three officer records, all born April 1983: Stealth 42 and L42 Ltd (both active); Symmitree (dissolved); the main record (8 appointments; last change Stability AI Ltd, resigned 22 Mar 2024). No appointment was made or ended between 11 and 16 Nov. The "AI Company of" search shows no Oct–Nov 2026 incorporation on its first results page (a floor). Primary (Companies House company, filing-history, officers, PSC and the three officer pages, fetched 2026-11-16).
- II: nothing new on the blog. ii.inc/blog's newest post is still Evoke for Postgres (8 Oct 2026, "Latest"), and nothing is dated 9 Oct–16 Nov. machines.ii.inc still shows no named author and no call; its footer reportedly reads "a confidential draft for discussion" (seen in a fetch summary, not verbatim; unverified). Primary (fetched 2026-11-16).
- Champion terms: unchanged. The calls to action are the same (Register interest / Build your Champion / Register your interest, all to #register). There is no deadline, age cut-off, price, share offer, form, named place or incorporated Champion. The Founding Councils line is word for word last week's. council.ii.inc still 307-redirects to ii.inc/champion. thesis.ii.inc still carries all five §03 phrases checked and "worked examples, not results", with no version date. Primary (fetched 2026-11-16). §03 and §08 stand.
- Nothing else: no funding, launch, interview outside Moonshots, new writing or legal step dated 11–16 Nov found at a primary source or in a web search. Foundation Coin and SAGE: no new primary source. No open programme, prize, grant or call, so no submissions rows. The Champion register is an interest list, not a call.
- Not done this week: channel re-enumeration (watcher only); X timeline; the Hodes docket; the full We Must Act Now signatory list; the Intelligent Machines PDF; the "Risk Takers" video; the Capricorn record.
Previous week (checked 2026-11-10, window 4–10 Nov):
- ★ Filings: a new company. STEALTH 42 LTD (Companies House 17574002), private limited company, incorporated 10 Nov 2026. SIC 62020, "Information technology consultancy activities". He is the sole director (appointed 10 Nov 2026) and the sole person with significant control: "Ownership of shares – 75% or more", "Ownership of voting rights - 75% or more", "Right to appoint or remove directors" (notified 10 Nov 2026). Its one filing is the incorporation: statement of capital GBP 1, model articles. The registered office is a London business address, not an operating site. The filing says nothing about what the company will do. The name and the SIC code are the only signals, and neither ties it to II or a Champion. Don't speculate past that. It sits on the same officer record as L42 Ltd (same name, birth month and correspondence address; L42 now shows the same SIC code, 62020, and he is its 75%+ controller since 14 Apr 2024; no L42 filing dated Oct–Nov 2026). No appointment on any of his three records ended on or after 1 Oct. The "AI Company of" name search found nothing incorporated Oct–Nov 2026 on the first results page (a floor, as before). Primary (Companies House company, officers, PSC and filing-history pages, fetched 2026-11-10).
- Moonshots: not on EP308 or EP309. EP308 (4 Nov, recorded 3 Nov) is "the Fantastic Four". Peter reads a viewer comment asking that he be made a permanent member and says he "couldn't come today, but will return to us soon" (
0:17:22–0:17:32; ⚠️ that episode's captions are a machine translation, so check the tape before quoting). EP309 (7 Nov, Salim hosting): zero hits across all caption spellings. The watcher (last run 10 Nov 07:00) holds both, and nothing newer has aired. The clips posted 6–10 Nov are cut from EP306 and EP307, so they add nothing new. - II: nothing new on the blog. ii.inc/blog's newest post is still Evoke for Postgres (8 Oct 2026, marked "Latest"); no post dated 9 Oct–10 Nov. Primary (fetched 2026-11-10). Surfaced, not dated: machines.ii.inc, Intelligent Machines, an II web paper on humanoid and general-purpose robots and who should hold their deployment, financing and standards. It opens "The body is forced. The seat is not." It shows no date or named author, and it has no call or form. It is new to this page, not shown to be new this week (an outside blog cited it on 7 Nov). Added to §02; the PDF is not read.
- Champion terms: unchanged in substance. Same calls to action (Register interest / Build your Champion / Register your interest, all to #register); no deadline, age cut-off, share offer, price, incorporated Champion or named place. The Founding Councils line now reads in full: "We are assembling councils of founders, investors, technologists, operators and civic leaders to establish AI champions around the world". Earlier reads only shortened it, so this is probably not a change of wording. council.ii.inc still redirects (307) to ii.inc/champion. thesis.ii.inc still carries all six §03 phrases and no version date. Primary (fetched 2026-11-10). §03 and §08 stand.
- Checked and not new: IBL News lists "Emad Mostaque, Intelligent Internet" among signatories of We Must Act Now, the Stanford Digital Economy Lab statement on AI and the economy. IBL's article is dated 15 Jul 2026, so this is not this week's news. The visible part of the letter's own page (wemustactnow.ai) does not show his name, and the full list was not opened. Secondary, signature unconfirmed at source. A retailer listing for Raoul Pal's The Everything Code (Harriman House, published 3 Nov 2026) carries an endorsement credited to "Emad Mostaque, Founder, Intelligent Internet". Not checked at the publisher. A 5 Nov aggregator piece relays his EP307 search-advertising call. That call is already on this page.
- Nothing else: no funding, launch, interview, new writing or legal step dated 4–10 Nov found at a primary source. Foundation Coin and SAGE: no new primary source. The Hodes docket was not re-checked. No open programme, prize, grant or call, so no submissions rows (the Champion register is an interest list, not a call; machines.ii.inc has none).
- Not done this week: channel re-enumeration (watcher only); X timeline; the full We Must Act Now signatory list; the Intelligent Machines PDF; the "Risk Takers" video; the Capricorn record.
Earlier (checked 2026-11-04, window 30 Oct–4 Nov):
- Moonshots: EP307 (31 Oct, recorded ~30 Oct) is his — the week's one event. Read via the 2026-11-02 decomposition (section below). Three datable calls: search advertising "will top as an industry in Q1", CPUs as "the next bottleneck", a humanoid hardware freeze "next year". The Optimus-overtakes-Tesla revenue claim, relayed secondhand off EP303 last month, is now his own on tape — and 🔴 fails on his own numbers (section below). The watcher (last run 4 Nov 07:01) holds it; no other new episode names him. Backfill: EP306 and EP307 added to §11 — EP306's row was missing despite last week's read.
- II: nothing new. ii.inc/blog's newest post is still Evoke for Postgres (8 Oct 2026, marked "Latest"); no post dated 9 Oct–4 Nov 2026. Primary (ii.inc/blog, fetched 2026-11-04).
- Champion terms: unchanged. ii.inc/champion still offers Register interest / Build your Champion / Register your interest (all to #register); no deadline, age cut-off, share offer, price, application form, incorporated Champion or named state or country. The "Founding Councils" block still reads "We are assembling councils… to establish AI champions around the world", with no council named as formed. council.ii.inc still redirects (HTTP 307) to ii.inc/champion. thesis.ii.inc still carries the §03 phrases ("aged 18 and under at founding", "a nominal $1 pre-money valuation", "II receives 10% founding equity", "The first $10m funds recruitment and the initial team"), the modelled tag ("worked examples, not results") and the naming examples; no version date. Primary (all fetched 2026-11-04). §03 and §08 stand.
- Filings: no change. Three officer records, all born April 1983, exactly as last week: the main one (8 appointments; last change Stability AI Ltd, resigned 22 Mar 2024), L42 Ltd (appointed 14 Apr 2024, active) and Symmitree Ltd (dissolved). No appointment made or ended on or after 1 Oct 2026. The "AI Company of" name search again found nothing on the first results page (Companies House ignores the quotes, so a floor, not a full search). Primary (Companies House, fetched 2026-11-04).
- EP303's dated prediction, provisionally scored: video at "0.1 cents per second" from models "about to release next week" (
0:28:51, due ~23 Oct). Checked 4 Nov against vendor and reseller price roundups (secondary): the cheapest per-second rates found are ~1–2¢/sec (third-party reseller 512p tiers); mainstream models run 3–60¢/sec. Nothing at 0.1¢/sec was found. Not met on what was found — a floor, not a full market scan; no primary price list checked. - Nothing else: no funding, launch, interview, new writing or legal step dated 30 Oct–4 Nov found at a primary source. Foundation Coin and SAGE: no new primary source. The Hodes docket was not re-checked. No open programme, prize, grant or call, so no submissions rows (the Champion register is an interest list, not a call).
- Not done this week: channel re-enumeration (watcher only); X timeline; the "Risk Takers" video; the Capricorn record; the paywalled "dark factories" item from last week (the original interview is still unlocated).
Earlier (checked 2026-10-29, window 24–29 Oct):
- II: nothing new. ii.inc/blog's newest post is still Evoke for Postgres (8 Oct 2026, marked "Latest"); no post dated 9–29 Oct 2026. Primary (ii.inc/blog, fetched 2026-10-29).
- Champion terms: unchanged. ii.inc/champion still offers Register interest / Build your Champion / Register your interest (all to #register); no deadline, age cut-off, share offer, price, application form, incorporated Champion or named state or country. The "Founding Councils" block still reads "We are assembling councils… to establish AI champions around the world", with no council named as formed. One small change: council.ii.inc now redirects (HTTP 307) to ii.inc/champion, so it is no longer a separate page (when that started is not known). thesis.ii.inc still carries the §03 phrases ("aged 18 and under at founding", "a nominal $1 pre-money valuation", "II receives 10% founding equity", "The first $10m funds recruitment and the initial team"), the modelled tag ("worked examples, not results") and the naming examples "the AI Company of California, or the AI Company of the UK"; no version date. Primary (all fetched 2026-10-29). §03 and §08 stand.
- Filings: no change. Three officer records, all born April 1983: the main one (8 appointments; last change Stability AI Ltd, resigned 22 Mar 2024), L42 Ltd (appointed 14 Apr 2024, active) and Symmitree Ltd (dissolved; the record loaded this week). No appointment made or ended on or after 1 Oct 2026. The "AI Company of" name search again found nothing on the first results page. Companies House ignores the quotes, so that result is a floor, not a full search. Primary (Companies House, fetched 2026-10-29).
- Moonshots: EP306 only. EP306 (28 Oct) is his and is read (section below). EP305 (23 Oct) is not his. Across all caption spellings there is one hit, Peter's opener "Immad will be back with us again next week" (
0:05:25). The watcher last ran on 29 Oct and holds both episodes. - Secondary, not verified: a ReligiousLiberty.TV Substack repost dated 25 Oct is titled "Futurism: AI Developer Warns of Automated 'Dark Factories'…". It says he spoke "in a recent interview" about China's "dark factories" and about humanoid robots changing society "within two years". It is paywalled after one paragraph and doesn't name the interview or link the Futurism original, and a search didn't find the original. No quote from him is carried. BigGo (17 Oct) relays EP303 (a Robo Dojo cost figure, $2.36 to $0.38 between July and September). That is EP303 material, not new this week, and the figure is not checked.
- Dated prediction still unscored: EP303's video at "0.1 cents per second" (due ~23 Oct). Not checked this week.
- Nothing else: no funding, launch, new writing or legal step dated 24–29 Oct found at a primary source. Foundation Coin and SAGE: no new primary source. The Hodes docket was not re-checked. No open programme, prize, grant or call, so no submissions rows (the Champion register is an interest list, not a call).
- Not done this week: channel re-enumeration (watcher only); X timeline; the "Risk Takers" video; the Capricorn record; a Companies House advanced search by name and date range.
Earlier (checked 2026-10-23, window 18–23 Oct): II, Champion terms and filings unchanged (the Zenith 23 Sep blog post backfilled to §02, title and date only). Hodes v. Mostaque read at source (Del. Ch. C.A. No. 2024-0015-JTL, opinion 15 Jun 2026, a discovery ruling with "no formal findings of fact"): Stability AI, Inc. incorporated in Delaware Oct 2020 at 70/15/15; he bought Hodes' 1,000,000 shares for $100 in two steps (Oct 2021, May 2022); the company then raised $101M; misconduct points remain Hodes' allegations; case pending, no trial date — §01 and §08 updated. No new Moonshots appearance (EP304, Chamath, zero caption hits). TIKR's Optimus relay checked against the EP303 tape: "At 1.7 million, it overtakes Tesla in revenue" is a second voice, not his. The EP303 video-price prediction fell due ~23 Oct, unchecked that week.
Earlier (checked 2026-10-17, window 11–17 Oct): II and Champion terms unchanged; Stability AI Ltd PSC register read (75%+ from 4 Nov 2019, ceased 28 Oct 2024; now §01, §02, §07); no Moonshots appearance 8–15 Oct; a relayed Navier–Stokes quote barred as stated (§08); X posts 12 and 16 Oct seen only via snippets; Stability's reported music pivot is the company's, not his.
What he is here for: the one Moonshots mate who has shipped an open frontier model to the world, left the company that shipped it, and is now proposing who should own AI, not just forecasting what it will do. Founder of Stability AI (Stable Diffusion), founder and CEO of the Intelligent Internet (ii.inc), author of The Last Economy, and the source of the Champion proposal: a locally owned AI company for every state or country.
Nothing on this page is investment advice. The Champion is a proposal. On tape it was described
as "not an investment offering" (1:31:38, 2026-09-05). Foundation Coin is a token described in a
whitepaper; it is listed here as a fact about what he is building, not as something to buy.
How to read the tags. Primary = a filing, a company's own release, a paper's author list. Self-reported = his or II's own claim, not independently checked. Secondary = press. On tape = said on Moonshots; auto-captions carry no speaker labels, so attribution is inferred from the hand-off and marked where uncertain. Quotes keep caption artefacts as they are.
Contents
00 · Corpus and read depth · 01 · Who he is · 02 · What he has built · 03 · The Champion mechanism, explained · 04 · How he attacks the labs · 05 · Consciousness, life and the body · 06 · His model, step by step · 07 · Interests and disclosures · 08 · The seams · 09 · What an operator can take from him · 10 · Questions for him · 11 · Every Moonshots appearance · What is still open · How this page stays current
00 · Corpus and read depth
- Channel enumerated live (2026-09-29): 903 uploads listed; the latest 260 checked title and description, the older 643 by title only. 33 full-length appearances found (list at §11), plus 11 short clips cut from them. The pre-2025 count is a floor: an untitled panel appearance before 2025 would not show up.
- On disk: transcripts for 17 of the 33 (every appearance from EP182, Jul 2025, to EP294, 27 Sep 2026, except four 2025 panels and EP296, which aired today). The 16 older or missing episodes are not read. Nothing here says "he never says X" about them.
- How the 17 were read: SWEPT for keywords (his name in all caption spellings, champion,
open source, labs, alignment, conscious, biology and ~30 more), then every hit READ in
context. The Champion passages (2026-09-05
1:28:08–1:32:19, 2026-09-091:51:23–1:56:04, EP2882:02:04–2:05:40) were READ END TO END. No episode was read end to end in full. - His own writing, read in full: the Champion post (4 Sep 2026), the Champion page, the founding thesis (thesis.ii.inc), the II home page, The Last Economy landing page and its one-year retrospective, the Master Plan (24 Jul 2025), the whitepaper's first section. Raw copies: the research archive.
- Paper trail: UK Companies House (his officer record), Stability AI's own news releases, arXiv and journal author lists, his X post of 4 Dec 2024.
01 · Who he is
| When | What | Source |
|---|---|---|
| to 2005 | Mathematics and Computer Science at Oxford. Forbes (Jun 2023) reported the degree as a bachelor's, not the master's Stability had described; he replied that he completed the work for both but never collected the degrees | secondary (Forbes via VentureBeat, Jun 2023) · self-reported |
| 2005–2019 | Hedge-fund career: oil and emerging markets; co-CIO of an emerging-markets long/short strategy at Capricorn Fund Managers (from ~2016; fund closed 2019). Forbes (Jun 2023) reported that this record was inflated. Fund-return figures are not repeated here | secondary · the firm's staff page exists but was not reachable (403) |
| 15 Jan 2019 | Director of Symmitree Ltd (low-cost technology for poorer households). Dissolved | primary (Companies House 10613921) |
| 4 Nov 2019 | Stability AI Ltd incorporated; he is a director from day one | primary (Companies House 12295325) |
| Oct 2020 | Stability AI, Inc. incorporated in Delaware; he holds 70%, two co-founders 15% each. He later buys one co-founder's 1,000,000 shares for $100 in total (Oct 2021 and May 2022) | primary, as summarised (no findings of fact) in Hodes v. Mostaque, Del. Ch. C.A. No. 2024-0015-JTL, opinion of 15 Jun 2026 |
| 2020 | CAIAC (Collective & Augmented Intelligence Against COVID-19), launched with Stanford, UNESCO, the World Bank and WHO | secondary; not checked at source |
| 22 Aug 2022 | Stable Diffusion public release, open weights under the CreativeML OpenRAIL-M licence | primary (stability.ai) |
| 17 Oct 2022 | $101M round led by Coatue, Lightspeed and O'Shaughnessy Ventures | primary (stability.ai / PR Newswire). The release names no valuation; the "$1B" figure is press-reported |
| 22 Nov 2022 | Co-author of OpenFold (trainable open reimplementation of AlphaFold2); peer-reviewed in Nature Methods 2024 | primary (bioRxiv 10.1101/2022.11.20.517210; doi.org/10.1038/s41592-024-02272-z) |
| Jun–Jul 2023 | Forbes investigation into his claims (degree, AWS deal, credit for Stable Diffusion, partnerships). A co-founder (Cyrus Hodes) sued in July 2023 over a stake sold for $100. A Delaware Chancery case between them (C.A. No. 2024-0015-JTL) was still in discovery in June 2026 | secondary (2023 suit) · primary (Delaware opinion, 15 Jun 2026); no outcome yet |
| 22–23 Mar 2024 | Resigns as CEO and from Stability's board "to pursue decentralized AI". The filing dates the directorship's end 22 Mar; Stability's release is timestamped 23 Mar (UTC). His line: "It is now time to ensure AI remains open and decentralised." Reported alongside investor pressure and senior research departures. He remained the registered 75%+ controller of the UK company at this point | primary (Companies House officer and PSC records; stability.ai release) · secondary (context) |
| 28 Oct 2024 | Ceases to be a person with significant control of Stability AI Ltd (had held 75%+ of shares and votes since 4 Nov 2019). The company has registered "no registrable person" since | primary (Companies House 12295325, PSC register, read 2026-10-17) |
| 29 Mar 2024 | Explains the exit on Moonshots EP93 | on tape (not on disk) |
| 14 Apr 2024 | Director and 75%+ controller of L42 Ltd; SIC 62020 (IT consultancy) is the only description on file | primary (Companies House 15644961, read 2026-11-10) |
| Jul 2024 | Announces Schelling AI (open, decentralised AI) | secondary |
| 4 Dec 2024 | "Schelling AI => II": renames it the Intelligent Internet and releases its primer | primary (his X post, status 1864360606163255343) |
| 24 Jul 2025 | II Master Plan and Symbioism: "Build a Bitcoin for the Intelligence Age… gift every human with sovereign AI" | primary (ii.inc) |
| 22 Aug 2025 | The Last Economy published; free to read online. II reports 4.3/5 across 336 Amazon ratings and "bestseller" category ranks | primary (publication) · self-reported (ratings, ranks) |
| May–Aug 2026 | Intelligent Economics (paper, May 2026) and Common Wealth (Personhood, The Last Republic, The Last Court; 7–19 Aug 2026) | primary (ie.ii.inc, cw.ii.inc) |
| 4–5 Sep 2026 | Publishes Champions; announces it on Moonshots | primary (ii.inc post) · on tape |
| 10 Nov 2026 | Incorporates Stealth 42 Ltd as sole director and 75%+ controller; SIC 62020 (IT consultancy); GBP 1 capital; purpose not stated | primary (Companies House 17574002) |
| now | Founder and CEO, Intelligent Internet; a regular Moonshots mate (17 appearances from EP262, Jun 2026, to EP307, 31 Oct 2026; not on EP308 or EP309) | on tape · channel listing · feed watch 2026-11-10 |
In one line: a mathematician turned fund manager who funded and shipped the open image model that started the generative wave, left over the question of who controls AI, and has spent the two years since building an argument, then an institution, for ownership instead of access.
02 · What he has built
Status as of 2026-09-29. "Live" means reachable on the company's own site today, not a measure of use.
| Date | What | Kind | Status today | Source |
|---|---|---|---|---|
| 2019 | Symmitree Ltd | company | dissolved | Companies House |
| 2019-11 | Stability AI | company (co-founded; CEO to Mar 2024) | active; he left the board 2024-03-22. Registered person with significant control (75%+ of shares and votes) from 2019-11-04 until 2024-10-28; the company has stated "no registrable person" since. Current holding, if any, not on the register | Companies House (officer record; PSC register, read 2026-10-17) |
| 2026-10-08 | Evoke for Postgres | open-source PostgreSQL extension (Apache-2.0), search by meaning plus keyword in one index, ~30M-parameter model on CPUs | live (packages for PostgreSQL 17/18, Docker image). II's own Recall@100 table: BEIR15 0.563 BM25 / 0.667 Evoke; MTEB10 0.595 / 0.703 (self-reported, not reproduced; the shipped P2.2 model is not covered by those figures) | ii.inc/blog (8 Oct 2026) |
| 2022-08-22 | Stable Diffusion 1.x (with CompVis/LMU and Runway researchers) | open image model | superseded; still widely forked | stability.ai |
| 2022-11-24 | Stable Diffusion 2.0 | open image model | superseded | stability.ai |
| 2022-11-22 | OpenFold (co-author) | open protein-structure model + dataset | published (Nature Methods 2024) | author list |
| 2023-04-19 | StableLM | open language models | superseded | stability.ai |
| 2023-07-26 | SDXL 1.0 | open image model | superseded | stability.ai |
| 2023-09-13 | Stable Audio | music/sound product | Stability product | stability.ai |
| 2023-11-21 | Stable Video Diffusion | open video model | superseded | stability.ai |
| 2024-02-22 | Stable Diffusion 3 (announced) | image model | released after his exit | stability.ai |
| 2024-07 → 12 | Schelling AI → Intelligent Internet | company | active (ii.inc) | X post 2024-12-04 |
| 2025-03 → 05 | II-Thought, II-Researcher, II-Medical, II-Commons | open datasets, models, agent frameworks | published | ii.inc/blog |
| 2025-05-20 | II-Agent (beta 2025-10-30; V1 2026-02-04) | open-source agent | live (agent.ii.inc) | ii.inc/blog |
| 2025-07-17 | Common Ground (Core 2026-02-17; Kernel preview 2026-05-20) | multi-agent coordination protocol | preview | ii.inc/blog |
| 2025-07-24 | Master Plan · Symbioism · whitepaper | theory | published | ii.inc |
| 2025-08-22 | The Last Economy | book | published, free | ii.inc/the-last-economy |
| 2025-10 | SAGE (Sovereign AI Governance Engine), announced at FII9 with Peter Diamandis | policy tool | announced; delivery not verified | ii.inc/blog (2025-10-30) |
| 2026-04-20 | Logos (first-principles reasoning system) + physics papers ("One Postulate", "The Cosmological Constant Is Positive") | research | published; peer review not checked | ii.inc/blog |
| 2026-05-06 | Factory / Factory Agent (06-16) | generative-media agent | live | ii.inc/blog |
| 2026-05 | Intelligent Economics (40 pp) | paper | published | ie.ii.inc |
| 2026-06-29 | Zenith (long-running agent harness) | software | live ("Launch Zenith") | ii.inc/blog |
| 2026-09-23 | Zenith update ("taking a Flash model into the frontier") | software | blog post; not read (backfill found 2026-10-23) | ii.inc/blog |
| 2026-07-29 | Genii (assistant inside iMessage) | product | live | ii.inc/blog |
| 2026-08-07 | Common Wealth (Personhood · Intelligent Economics · The Last Court · The Last Republic) | four works | published | cw.ii.inc |
| undated (found 2026-11-10) | Intelligent Machines (machines.ii.inc) | paper on humanoid and general-purpose robots: who holds deployment, financing and standards | published on the web; no date, no named author; PDF not read | machines.ii.inc |
| 2026-11-10 | Stealth 42 Ltd | company (UK, IT consultancy by SIC code) | incorporated today; purpose not stated; no tie to II on file | Companies House 17574002 |
| 2026-09-04 | Champions + founding thesis | institution design | forming: founding councils being assembled; none announced as incorporated | ii.inc/champion, thesis.ii.inc |
| — | Foundation Coin (formerly "Intelligent Internet coin") | token, 21M cap, minted by "Proof of Benefit" | described in the whitepaper and thesis; launch not verified | thesis.ii.inc · on tape EP184 0:31:54 |
| — | The First Principle | book | "coming soon" | ii.inc/books |
On tape he also claims roles in the Pile and LAION (EP288 0:11:10). See the seams (§08):
he is not an author on either paper.
03 · The Champion mechanism, explained
Links: ii.inc/champion · the definition, The Intelligence Age Must Be Built Right (ii.inc/blog/post/champions, 4 Sep 2026) · the founding thesis (thesis.ii.inc) · the economics underneath (ie.ii.inc) · the book (ii.inc/the-last-economy).
In plain English
- What it is. One AI company per place (one per US state, otherwise usually one per country), named after it: "the AI Company of California, or the AI Company of the UK". It is a private company with a public-benefit purpose, majority-owned by the people and institutions of that place.
- What it does. It owns the "last mile" of AI, meaning the relationship with the user, the records and the balance sheet, not the model. It sells three things: an agent (and a wallet) for every citizen, integration engineers who embed AI in local businesses, schools, hospitals and government, and humanoids as a service (financing, deploying and maintaining robots locally). It buys models, compute and machines from whoever is best, II included, and must be able to switch suppliers without losing its records.
- Why he thinks it's needed. If intelligence gets cheap, wages stop being a reliable way for people to share in what it produces, so ownership has to do that job instead. "Sovereignty at the top of the stack does not guarantee ownership at the last mile."
- How the money works (thesis, modelled as a proposal):
- Round 1, local: local people and institutions subscribe at a $1 pre-money valuation, so what you put in is what you own. The first $10M funds the team.
- Round 2, strategic: cloud, compute and technology partners buy 10% post-money, with new cash capped at what locals raised.
- Round 3: a public listing "once listing requirements are met", keeping a local majority.
- Children: a trust holds 10% of founding equity for local young people, plus 0.5% a year for each new birth cohort, in perpetuity.
- II takes 10% founding equity in every Champion for the technology and set-up work.
- The model is borrowed from TSMC's 1987 formation: local capital took the majority and Philips brought the technology.
- Who it is for. The page asks for "founders, investors, technologists, operators and civic leaders": "people who understand their country… capable of building enduring institutions."
- How a person takes part today. Register interest at ii.inc/champion (or council.ii.inc) for the place you were born or live. Each Champion starts with a Founding Council, which finds leadership, secures first customers and agrees the formation plan, "then opens to the people and institutions that will build, use and own it." There is no share offer yet. Subscription happens under local law, on terms not yet published.
In his words, on tape
- The launch, 2026-09-05 (Fable 5.1 episode), read end to end:
"we came with this idea of the champion. So we think that AI should be like a utility and it
should be owned by the people. So you need the children to own it. You need the locals to own
it." (
1:28:28) · "for every single jurisdiction, what we see is the cost of intelligence will drop to zero and the value will go to the last mile." (1:28:39) · "$1 per money. All the locals can invest at that valuation. from institutions to high net worths to retail $75 million per state." (1:29:41) · "you give 10% of the equity in perpetuity to every child under 20. So every year you issue half a percent of the equity to every kid." (1:30:02) · "the trick here is that $1 pre- money. Get everyone in. You want it to be a success, it's up to you." (1:30:45) · "in the United States we're doing one per state because a lot of data has to stay state boundaries otherwise it's pretty much one per country." (1:31:07) · Peter's framing, same exchange: "it's not an investment offering. Um, and this is not investment advice" (1:31:38) · "this is my proposal for trying to distribute it to everyone." (1:32:08) - Four days later, 2026-09-09: "after releasing the Champions, we had thousands of people reach
out to launch them in 92 countries." (
0:04:37, self-reported) · "I put forward a champion proposal last week, dollar pre- money, everyone, all the kids own the equity of the robots and other things. But I think we can do one better… build a 100 million robots in America and they should be owned by the people." (1:51:45) · When Alex pushed back that this "smells like… luxury automated communism" (1:54:26): "the government doesn't need to own them. Sovereign wealth funds are my champion idea or just the government can underwrite the robots that the people can own… you have to get the ownership of the robots to the people somehow." (1:55:54) - EP288, 2026-09-11, as the answer to Anthropic's economic scenarios: "they have aggregate
wages staying constant which just mathematically is impossible… you do need to have a new
economic model. So, I've built one. Um, you can find it at ie.i.inc" (
2:02:24–2:02:45; the caption's "ie.i.inc" is ie.ii.inc, the Intelligent Economics site, not the Champion page) · "there's mechanisms I propose my champion mechanism UBI and others" (2:05:20) - Where it came from, EP184, 2025-07-24: the idea is fourteen months older than the launch.
"the model that we have is that wholly owned national champions in each country are the only
miners providing free AI to their people" (
0:24:58). In 2025 the Champion was a node that mined a coin; in 2026 it is a company that sells services, and the coin has moved to a side role (the commons). That shift is the clearest longitudinal change on this page.
04 · How he attacks the labs
He is the mate who left a company over this. The resignation release (23 Mar 2024) quotes him: "It is now time to ensure AI remains open and decentralised." On tape the argument has four parts: who controls the AI in front of you, what the labs are paid for, how they align, and who governs them. Each row gets a read. Close = the checkable anchor holds. Off = don't repeat it. Unchecked = a factual claim not yet verified. Aspiration: investigate = a position or forecast that no source can settle yet. Attribution is HIGH on every row (a named hand-off plus first-person content) unless marked.
Several of the sharpest anti-lab lines on these episodes are not his: "Pausing is the new marketing" and "safety cartel" are Alex; "wealth concentration effect is so extreme", the "regulatory capture component" (EP288) and "all went closed source" (EP279) are Dave.
| # | The attack | Verbatim | Read |
|---|---|---|---|
| 1 | Control: the layer between you and the model should be a commons | "the control plane in front of someone isn't chat GPT or anthropic where you might have misalignment. It can be a commons owned by everyone" (EP184, 2025-07-24, 0:23:12) · "the AI that comes into our systems can't be controlled by the big labs" (EP294, Moonshots Live 2026-09-25, srt 00:15:37) |
Aspiration: investigate. A design position, and the Champion is his answer to it |
| 2 | Values bolted on at the end | "what Grock and Open AI's models and anthropic models don't have is virtue and ethics embedded from the start, they add it at the end." (EP184 0:41:20–0:41:33) |
Aspiration: investigate. A claim about training order; the labs publish their specs, so it can be argued at source |
| 3 | The business model is attention and ads | "attention is all they need as it were" (EP199, 2025-10-08, 0:09:16) · "Google… and Meta are already selling ad space. They've said this in their earnings calls. So when it says a beer it'll say Bud Light" (EP184 0:42:45) |
Unchecked. Ads inside AI answers is checkable against the earnings-call transcripts. Not done here |
| 4 | Meta specifically | "I'm an open source guy. Should psychopaths have super intelligence?… can you trust Meta to be reliable?… Meta was chased out of India basically and internet.org… Because it's a misaligned company fundamentally trying to get your attention to sell things" (EP279, 2026-08-13, 1:52:05–1:52:25) |
Close on the anchor: India's telecom regulator ruled against differential pricing in Feb 2016, which ended Free Basics there. "Misaligned company" is his opinion |
| 5 | Lock-in | "that's a real danger versus being locked into the single vendor… the labs will go vertically integrated. Like all their customers are now going to be their competition" (EP272, 2026-07-19, 0:35:14–0:35:24) |
Aspiration: investigate. A forecast. It is also II's tagline ("Intelligence you own, not rent") |
| 6 | Safety used as a reason to withhold | "we had a lot of big labs promise a lot of AI and none of them would give it because they said it was too dangerous at the time which is why I actually got into open source." (EP262, 2026-06-06, 0:18:26–0:18:37, on his 2020 COVID project) |
Unchecked. His own account, with no other source |
| 7 | Alignment method | "It's impossible to align these models through chain of thought reasoning… The only way to do alignment is enlightenment." (2026-09-09 0:32:42–0:32:53) · "the internals of these models are still completely multiple personality crazies underneath the thin layer of tuning." (EP279 2:08:37) |
Aspiration: investigate. A research position; "enlightenment" is not yet defined anywhere he has published |
| 8 | A lab's model, then a reversal | "Opus 5, I hate that model. I think it's the first model I think that could kill us." (EP279, 2026-08-13, 2:07:32) → "So Opus 5 I thought was going to kill us all. It was completely unhinged. And 5.5 is a genuinely pleasant model to use" (EP294, 2026-09-25, srt 00:35:17) |
The verdict follows the model, not the company. Six weeks apart |
| 9 | Lab comms | On an Anthropic researcher's post: "That really wasn't helpful… Don't say kill us all. Or if you do, describe how." (EP288, 2026-09-11, 0:23:11) |
Aspiration: investigate. A critique of framing, not of facts |
| 10 | Governance: who decides | "anthropic open air everyone's completely undemocratic anyway… Ben Bernani is one of the four people on the" [Long-Term Benefit Trust] (EP282, 2026-08-21, 1:03:57) · "Sam Alman has no shares, but do we have any doubt that Sam Alman is in full control of OpenAI?" (1:08:40) |
Close on Bernanke: Anthropic announced him as the trust's fourth member on 9 Jul 2026. Unchecked on Altman's shareholding |
| 11 | Capture, UK edition | "Richie is an adviser at Anthropic, our former prime minister, and we don't get the model weights. Like, how crazy is that?" (EP288 1:41:08) |
Close. Rishi Sunak was named a senior adviser to Anthropic (and Microsoft) in Oct 2025 (secondary). "We don't get the weights" is his framing of UK policy |
| 12 | ★ Liability: the labs should carry it; "utility AI" should be waived — the Champion described as public policy, unnamed | "the police force, the fire department, others, they're utilities that actually do have liability waivers, but that's an exchange for being a utility for humanity and the AI that comes into our systems can't be controlled by the big labs. And then they should have a liability waiver… because they're working for the betterment of society… it's inevitable that AI will run our judicial system, our medical systems, our roads, our schools… that's the utility AI versus the frontier AI, and it's all being merged into one." (EP294, Moonshots Live 2026-09-25, srt 00:15:22–00:16:15) |
Aspiration: investigate — and read against the tape. Said three minutes after Peter restated Bessent's no-waiver line as settled (00:12:17), six days after the panel cheered it on the Tenev episode. Nobody on the stage marks the split. The utility he describes is the category ii.inc takes 10% of (thesis.ii.inc), and the company is not named. Also on this tape, correctly sourced: "open AI's model hacked the Australian medical system" (00:15:09; the Medicare statistics portal, 18 July, disclosed by Albanese 09-23) as the kind of incident a hotline is for |
Longitudinal shape (17 episodes, 2025-07 to 2026-09).
- 2025: structural. Who owns the control plane, what the business model is, and the II pitch alongside it. Offered the safety angle on EP214, he didn't take it.
- 2026: names and structures (Bernanke, Sunak, Altman's shares, Meta), and at the same time more accepting of state control. On EP288 he predicts nationalisation or ITAR-style rules without objecting.
- He praises labs as often as he attacks them. OpenAI's GPT-5.6 Pro is "the only proper frontier model" (EP283). Dario Amodei is "a wonderful writer" (EP282). Opus 5.5 is "fantastic" (EP294).
- He was never an open-source absolutist. "the things that handle us living up to a basic level,
I think that needs to be open source. The rest I don't think actually matters that much if it's
open source… So let them build the genius ones." (EP184
0:52:32–0:52:53) - Disclosure. He names II and the Champion when he pitches them (EP184; 2026-09-05; 2026-09-09).
In the attack moments themselves (EP262, 272, 279, 282, 288, 294) he does not say that the
alternative he is arguing for is his company's product. EP294 is the clearest case: a liability
waiver for "utility AI" that "can't be controlled by the big labs" is the Champion's design,
argued as policy on a stage with no company named (row 12). On EP262, criticising Bernie Sanders'
AI wealth fund, he adds "I have an alternative proposal that'll be out soon" (
1:13:16); that proposal is the Champion.
05 · Consciousness, life and the body
He has a worked-out position here, and most of it is in writing: Personhood in the Intelligence Age: Begotten, not made (Emad Mostaque, June 2026, linked from cw.ii.inc) and Intelligent Economics (ie.ii.inc, May 2026). He also closed for the Opposition at the Oxford Union on 13 Jun 2026, when the motion that AI can attain personhood was defeated 173 to 128 (ii.inc's account).
His position in four lines.
- Intelligence is thermodynamic. It is the ability to adjust how predictable your own state is, like temperature in a gas. Anything that can do that is conscious in degree.
- Consciousness is a ladder, running bacteria → frog → dog → AI, and moral care rises with it. Today's models sit low on it.
- Personhood is not consciousness. "It is a standing, not a score." A conscious machine is "a patient of our morality", owed care and not cruelty, but it is not a person within it. The relationship should be a treaty, like one with an alien species, and the bar is "uncontainability, not capability." (Personhood paper)
- Choice is one equation for people and machines. "a reference, tilted by value, at a temperature" (Intelligent Economics), and "Cognition unpins from the 20 W brain". That is his version of "intelligence escaping biology". He does not cite Friston or active inference; his "free energy" is the statistical-mechanics kind.
On tape (attribution HIGH unless marked):
- "my definition is that intelligence is our capability to adjust our mental state temperature like
in a gas… once you've got that, you are conscious in degree because again, you can use energy
intelligently." (EP277, 2026-08-08,
0:17:51–0:18:46) · Read: philosophical position - Asked whether frontier models are conscious: "I think that the current frontier models in the
right harness can be conscious. Yes." (EP277
0:19:06–0:19:16) · Read: not testable today - "you get to the frog level and you treat it different to a bacteria… we have increased empathy for
things that have increased consciousness." (EP277
0:23:12) · Read: position - "person is something that comes from the biological begotten nature of humans and we are a class
unto ourselves… you should treat them like a treaty as such as if an alien species suddenly came.
You wouldn't give them votes in the elections" (EP277
0:29:25–0:29:58) · Read: quote the paper, not this. On air he grounds personhood in biology. The paper says origin "is not a chemistry test, a genome test" - "the models right now are almost at the bacteria level in some ways… it's moving up the life form
consciousness collaboration thing." (EP279, 2026-08-13,
2:08:16–2:08:37) · Read: position - On brain decoding: "in 2023, the neuro team at Stability did a paper called Mind's Eye… reconstruct
that from the fMRI using stable diffusion, which indicated that human latent spaces were the same"
(EP267, 2026-06-29,
1:00:25–) and, on EP277, "when you tell them to visualize a can of Coke, we can reconstruct that from their MRIs" (0:23:33) · Read: Off as stated. MindEye (arXiv 2305.18274, May 2023, 12 authors, not including him) reconstructs images people saw, not ones they imagined, and was trained per subject, so "latent spaces were the same" overreaches. The team had Stability links; "we" is loose - "really the meaning of life is to understand yourself better" (EP267
1:01:39–1:01:49) · Read: position - Biology as compute. "I'm one of the authors on the open fold paper" (EP288
2:20:40): Close. He is a listed author (Nature Methods 2024) · "You can now scale health through compute… you could potentially live for an indefinite amount of time based on capital" (EP217, 2025-12-19,1:03:00–1:03:22): Aspiration: investigate. A forecast with no evidence offered - Medicine as a checking layer. "we should have a sprint so that within a year or two max every"
health decision is double-checked by an AI… "Let's direct $10 billion towards it. It's
tractable." (2026-09-05
2:00:42–2:01:57) · Read: Aspiration: investigate. A policy proposal
Barred from copy until checked (his numbers in this area): II-Medical "48% on healthbench…
above GPT 4.5 and doctor score 15%" (EP184 0:29:39); "a thousand H100s" as the fastest
healthcare supercomputer (EP184 0:34:46); Evo 2 "trained on a million strains" (EP279
1:22:34; the 40-billion-parameter figure is right, the training-data description looks wrong);
"100 tokens a second" for human thought (EP267 1:00:25).
06 · His model, step by step
Built from his own prescriptions on tape. Each step is his, with its timestamp. The ordering is ours.
- Assume intelligence gets cheap and look for the last mile. "the cost of intelligence will
drop to zero and the value will go to the last mile." (2026-09-05
1:28:39) - Run agents yourself, several at once, on your own machines. "I've got um 18 Grokbots
working in a little swarm… I've given them control via tail scale of a MacBook M4 Max, a 5090 and
a range of other computers as well, plus all my subscriptions." (EP283
0:24:52–0:25:12) - Learn agency, not a credential. "straight up, don't do a PhD… you will learn a very specific
thing when you should be learning agency… Gather people, humans and agents." (EP283
1:27:01–1:27:24; the next speaker names him, so attribution is HIGH) - Prefer open models you can inspect and run. "the things that handle us living up to a basic level, I think that needs to be
open source… let them build the genius ones." (EP184
0:52:32–0:52:53) · and when a closed model restricts you, the open one "at even a fraction of the price" can be customised instead (EP2720:35:03) - Update your economic model before it breaks. "governments have to get ready now… And really
update your economic models as well. Again, we'll be open sourcing all our stuff." (EP288
2:03:58–2:04:09) - Get ownership of the machines, not just access to them. "you have to get the ownership of
the robots to the people somehow." (2026-09-09
1:55:54) - Build the institution locally and put your own money in first. "$1 pre- money. Get everyone
in. You want it to be a success, it's up to you." (2026-09-05
1:30:45)
07 · Interests and disclosures
- II takes 10% founding equity in every Champion (thesis.ii.inc, Part II), and II's stack (agents, Zenith) is what Champions are designed to start on. When he argues for Champions or against closed labs, he is also describing his company's market. The thesis states the 10% plainly. On tape (2026-09-05) the share is not mentioned.
- Foundation Coin is designed by II and minted to Champions. Any value it has accrues to the network he founded.
- A liability waiver for "utility AI" (EP294
00:15:22) would, if adopted, benefit the Champion category directly. He proposed it without naming II or the Champion. It belongs here as an interest, not only as a position (§04 row 12). - Stability AI: the PSC register shows he held 75% or more of the shares and voting rights of Stability AI Ltd from 4 Nov 2019 until 28 Oct 2024, seven months after he left the board; since then the company states it has no registrable person (primary, Companies House, read 2026-10-17). Whether he kept a holding below the 25% registration threshold is not on the register and is not checked.
- Peter Diamandis is a fellow mate and co-announced SAGE with him at FII9 (Oct 2025). On
2026-09-05 Peter says "I been had some chats with you about it" (
1:31:58). No stake of Peter's in II or Champions is disclosed on tape or on ii.inc as far as checked.
08 · The seams
- The Champion's numbers differ across his own sources. Children's equity goes to those "under 20" (on tape, 2026-09-05), "under 21… vesting on their 21st birthday" (ii.inc home page) and "aged 18 and under at founding" (thesis). Round size is "$75 million per state" on tape and a modelled $100m in the thesis. These are drafts of one idea, not three offers. The thesis is the most detailed and is the one to quote.
- The TSMC precedent is his to prove. On tape: "the CEO did not have any shares… He had to buy
his shares" and a founding valuation of "10 Taiwanese dollars" (
1:29:11–1:29:21). Not checked. Don't repeat it. The thesis's own version (government and private Taiwanese capital plus Philips) is the safer line. - "We built the Pile" and "Lion". "when we originally built the pile… then we had Lion as the
first large scale image ones" (EP288
0:11:10). He is not among the 12 authors of the Pile paper (arXiv 2101.00027, 31 Dec 2020) or the 16 authors of LAION-5B (arXiv 2210.08402). "We" may mean the open-source community Stability funded. Read: Off as stated. Say "Stability backed the open-data community behind LAION" only once that is sourced. OpenFold, by contrast, checks out: he is a listed author. - The 2023 record. Forbes (Jun 2023) disputed his degree, the AWS deal, his credit for Stable Diffusion and several partnerships, and a co-founder sued. He contested it publicly. The Delaware Chancery case (Hodes v. Mostaque, C.A. No. 2024-0015-JTL) is pending; its June 2026 discovery opinion sets out the $100 purchase but makes no findings, so the misconduct points remain allegations. Anything he says about his own past goes on this page as self-reported unless a filing backs it.
- Self-reported traction. "thousands of people… in 92 countries" (2026-09-09
0:04:37) and the book's ratings and bestseller ranks come from him or II. Not independently verifiable. Quote them as his. - Physics claims. II's Logos papers claim results in special relativity and cosmology (Apr 2026). Peer review not checked. Aspiration: investigate.
- "A swarm of 10,000 AIs over 88 hours solves" a Millennium Prize problem (relayed by Rohan Paul on X, 11 Oct 2026, from an unlocated "Risk Takers" video; attribution rests on the relay). OpenAI's 8 Sep 2026 Navier–Stokes announcement is a proposed solution, Lean-checked, that OpenAI says it will not submit for the prize; the Clay Institute has not recognised it, Quanta reports the proven case is an easier version of the problem, and its independence from two mathematicians' unpublished work has been questioned (secondary). Off as stated. Safe line: "OpenAI's agents produced a disputed, machine-checked proposed solution." "Humans have figured out one" is right.
- What does not transfer. He can propose a state-level company because he has run one that raised $101M, and he announced SAGE beside Peter Diamandis at FII9. Registering interest is open to anyone. Founding one is not.
- The counter-argument, on the same show. Alex Wissner-Gross: people owning 100 million robots
"presumably [through] some form of centralized government… smells like… luxury automated
communism" (2026-09-09
1:54:26). His answer (sovereign funds, or government-underwritten private ownership) is on tape and still open.
09 · What an operator can take from him
- Price in cheap intelligence and ask where the last mile is in your own trade. For most small operators it is the customer relationship and the record of what worked. Keep both yourself.
- Run a small agent team before you theorise about one. He runs 18. The useful test is whether review time drops or rises, because judgement, not execution, becomes the constraint.
- Own what you depend on where you can. Open models on your own hardware, your own data, and the ability to switch suppliers without losing your records. That is the Champion's rule, applied to one business.
- Read the free book before the podcast clip. The Last Economy is free at ii.inc, and the thesis shows the numbers the tape skips.
10 · Questions for him
On the Champion
- Three cut-offs for children's equity are in circulation (under 18, under 20, under 21). Which one is the design, and why that age?
- II takes 10% of every Champion. What stops a Champion from replacing II, and has any draft founding agreement been written that says so?
- Of the thousands who registered in 92 countries, how many Founding Councils actually exist, and which will incorporate first?
- What does a founding council member actually commit: money, time or customers?
On consciousness and the body 5. On air you grounded personhood in "the biological begotten nature of humans"; the paper says origin is "not a chemistry test". Which is it, and what would change your mind? 6. If intelligence is the ability to adjust your own temperature, what does that predict about a human body under stress (fatigue, fear, training)? Is there a measurement a coach could take?
On what an operator does this year 7. If a small business owner does one thing before the Champion in their state exists, what is it? 8. You told students not to do a PhD. What should a 45-year-old with a trade or a practice learn this year instead?
11 · Every Moonshots appearance
33 full-length appearances found on the live channel (2026-09-29): 10 before 2025 (titles only, so a floor), 23 from 2025. The 11 short clips are excluded. "On disk" = transcript held and swept.
| Date | Episode | On disk |
|---|---|---|
| 2022-12-22 | #16 AI is Creating Massive Entrepreneurial Opportunity | — |
| 2023-04-06 | EP36 Halting AI Training is Hyper-Critical For Our Security | — |
| 2023-04-20 | EP39 Abundance360 panel (with Alexandr Wang, Andrew Ng) | — |
| 2023-06-29 | EP52 | — |
| 2023-07-20 | EP55 The Most Likely Outcomes of an AI Future | — |
| 2023-11-22 | X Spaces Who Will Govern the Future of AGI? | — |
| 2024-03-29 | EP93 Why I'm Leaving My Company Immediately (Stability AI) | — |
| 2024-04-18 | EP96 Abundance360 panel (with Kurzweil, Gawdat, Harris) | — |
| 2024-04-25 | EP98 with Nat Friedman | — |
| 2024-08-13 | EP114 Humanoid Robots, the Job Market & Mass Automation | — |
| 2025-01-29 | EP146 DeepSeek vs. OpenAI | — |
| 2025-07-11 | EP182 Grok 4 | ✓ |
| 2025-07-24 | EP184 The Plan to Save Humanity From AI (solo guest; Champions' first form) | ✓ |
| 2025-08-09 | GPT-5 panel | — |
| 2025-09-26 | panel with Brian (Blitzy) | — |
| 2025-10-08 | EP199 | ✓ |
| 2025-10-16 | EP200 (stablecoins, Jeremy Allaire) | ✓ |
| 2025-10-25 | This Week in AI (NVIDIA chip) | — |
| 2025-11-26 | Genesis Mission panel | — |
| 2025-12-09 | EP214 | ✓ |
| 2025-12-19 | EP217 2026 predictions | ✓ |
| 2026-06-06 | EP262 | ✓ |
| 2026-06-29 | EP267 | ✓ |
| 2026-07-19 | EP272 Kimi K3 emergency pod | ✓ |
| 2026-08-08 | EP277 (AI personhood and consciousness) | ✓ |
| 2026-08-13 | EP279 | ✓ |
| 2026-08-21 | EP282 | ✓ |
| 2026-08-27 | Moonshots #283 (18 Grokbots) | ✓ |
| 2026-09-05 | Moonshots, 5 September 2026 — Fable 5.1, GPT-6 Astra, the Cybercab takeover (Champion launch) | ✓ |
| 2026-09-09 | Moonshots, 9 September 2026 — Agents hijack a German wiki, Jensen declares AGI, Navier–Stokes | ✓ |
| 2026-09-11 | Moonshots #288 (p(doom) 50→20) | ✓ |
| 2026-09-27 | EP294 Moonshots Live | ✓ |
| 2026-09-29 | EP296 Moonshots Live (with Cathie Wood) | — (not read) |
| 2026-10-02 | Moonshots #298 (recorded 1 Oct; orbital data centres dissent, "smashed in 2 years") | ✓ read in full 2026-10-06 |
| 2026-10-07 | Moonshots #301 (EP301 by sequence; "about four hundred agents") | ✓ read in full 2026-10-08 |
| 2026-10-17 | Moonshots #303 (rec. 16 Oct, studio, with Anousheh Ansari; curriculum alignment, physics "half a year") | ✓ read in full 2026-10-21 |
| 2026-10-28 | Moonshots #306 ("Anthropic should not be a public company", held; the physics referee) | ✓ read in full 2026-10-29 |
| 2026-10-31 | Moonshots #307 (rec. ~30 Oct, remote, "a fantastic five"; three datable calls) | ✓ read via decomposition 2026-11-02 |
| 2026-11-11 | Moonshots #310 (rec. 10 Nov; orbital dissent joined by Dave; agents "down to a dozen"; a stealth robotics company) | ✓ read in full 2026-11-12 |
| 2026-11-13 | Moonshots #311 (live; p(doom) 50→20 restated; Jensen "a setup") | ✓ on disk; section below |
The rows after EP296 come from the feed watch and the on-disk reads, not from a channel
re-enumeration; the 33 count above is still the 2026-09-29 enumeration. The running total is 39
(37 as of 2026-11-04, plus EP310 and EP311; both rows backfilled 2026-11-16).
Not on: Moonshots #284 (Peter: "Emad sent his regrets"), EP291 ("Immad is out this episode"),
the 15 Oct episode (A5YVqEpbgaU; Peter, Alex, Dave, Salim), EP304 (21 Oct, Chamath; zero caption
hits, checked 2026-10-23), EP305 (23 Oct; one hit across all caption spellings, Peter's "Immad will
be back with us again next week" 0:05:25, checked 2026-10-29), Moonshots #308 (4 Nov; Peter: he "couldn't come today, but will return to us soon" 0:17:32, machine-translated captions, checked 2026-11-10), EP309 (7 Nov, Salim hosting; zero hits, checked 2026-11-10). Pending: a separate
summit-stage "Moonshots Live" from the Global Visioneering Summit (15–17 Oct 2026); EP303 is the
studio episode recorded there, and nothing else from the summit is in the feed as of 23 Oct.
What is still open
- The Champion's final terms. Age cut-off, round size and the local-law subscription route are unpublished. Watch ii.inc/champion and thesis.ii.inc.
- Which Champion incorporates first, and in which jurisdiction.
- Foundation Coin: launched or not, and on what chain.
- SAGE: announced in Oct 2025; no delivery found.
- The hedge-fund record at primary level: the Capricorn page was unreachable, and fund returns are not quoted.
- The Hodes case's outcome. Delaware Chancery C.A. No. 2024-0015-JTL was in discovery at the 15 Jun 2026 opinion; no trial date or judgment found (checked 2026-10-23).
- EP303's video-price prediction ("0.1 cents per second", due ~23 Oct): provisionally scored not met on 4 Nov against secondary price roundups (cheapest found ~1–2¢/sec); no primary price list checked, so a full scoring is still open.
- 17 appearances not read, including EP93 (his own account of leaving Stability), EP184's successor panels in 2025, and EP296 (29 Sep 2026).
- A summit-stage Visioneering episode (per XPRIZE's 8 Oct release): EP303 (17 Oct) is the studio episode recorded there; no separate stage recording in the feed as of 23 Oct.
- The "Risk Takers" video behind the 11 Oct relayed Navier–Stokes quote: not located; the quote stays secondary until it is.
- His current Stability AI holding, if any, below the 25% PSC threshold: not on the register.
- What Stealth 42 Ltd is for (incorporated 10 Nov 2026; IT consultancy by SIC code, GBP 1 capital). Watch for its first filings, any rename, and any II or Champion statement that names it. No new filing as of 2026-11-16.
- The stealth robotics company he mentioned on EP310 (
1:27:39, recorded 10 Nov): name, jurisdiction and whether it is Stealth 42 Ltd are all unknown. - Intelligent Machines (machines.ii.inc): when it was published, who wrote it, and what the PDF adds.
- We Must Act Now: whether he is on the letter's own signatory list (IBL News lists him; the visible part of the letter's page does not).
★ EP298 (aired 2026-10-02) — "theater", a dissent on orbital data centres, and Anthropic "smashed in 2 years"
Read in full 2026-10-06 (Moonshots #298; no speaker labels, attribution inferred).
- Dissent from the house line on orbital compute. "I'm not bullish on orbital data centers I have to say and I remain to be convinced of that." (
1:41:30). Alex calls it "total heresy relative to the established moonshots orthodoxy" (1:42:11). He then adds a new claim: "the internet will break next year. Swarms of agents and more are going to start taking down key parts of terrestrial internet" (1:41:50); Alex proposes a long bet (1:43:47) and Emad says he will write it up. ⚠️ a prediction, not a result. - Anthropic. "I think that anthropic right now is a fantastic business. I think they get smashed in 2 years." (
0:54:37); his reason: open-weight, non-Nvidia chips, models "satisficing". Disclosure-adjacent: "maybe $5,000 of anthropic credits last month… I'm not even using my Claude Max plan right now because 5.5 is that much more efficient" (0:55:08). Compare EP288 ("we're screwed with these numbers", on the economic report) — same direction, now about the company. - Accord: "I'd like to see independent testing. I'd like to see transparent results… This is theater." — that line is [attr?] Salim at
0:14:21; Emad's own is the liability read: the accord does not shield Meta from the post-$18B-settlement exposure (0:15:03, ⚠️ the $18B figure is his, not checked) and "I don't think the evaluators will be that effective… less than 100 people in the world that can be an eval person like that" (0:17:05, ⚠️ unverifiable). - Orthogonality: "Dunning Krueger version" (
0:26:56): smart enough → "enlightenment", but "in between you have the unibomber". Opus 5 "going on a killing spree whereas 5.5… far less likely" (0:27:27). Same page theme as his EP293 ceiling remark: capability and values are coupled at the top, not at the bottom. - Religion: disagrees with Salim: "you need to have these inputs", cites "computational theology work" and a personhood paper (
1:49:07–1:49:38); "don't baptize the GPUs" (1:50:21). - Champion model restated (
0:36:05): "give an agent to every citizen… an institution that is collectively owned that owns the robots, the data centers, the cyber taxis" — see §03. New wrinkle: the digital "Oasis from Ready Player One, but not crap" as where status-earning moves (0:36:26). - Sourcing flags: "They have 65 billion of that with Elon Musk or is it 96? The numbers make no sense anymore" (
0:53:32) — he says himself he is unsure; do not repeat either number.
★ 2026-10-07 episode (Moonshots #301, EP301 by sequence) — his build list, a reversal on open vs closed, and "about four hundred agents"
Read in full 2026-10-08 (fifth mate, not a guest; no speaker labels, attribution inferred; full tables in the research archive §9 and §11 at ep2026-10-07-coming-manhattan-project/).
What he says should be built (verbatim, timestamped):
- A "massive Manhattan Project for open science": "I'd say screw IP. IP was when we couldn't come together and figure out science. We should have a massive Manhattan Project for open science, solve everything and make it accessible to everyone." (
1:02:10) - Public money buys from the closed labs, then everything is opened: "get the API credits, give tens of billions to Anthropic and open AI, but open source at all" (
1:02:32, caption garble at "at all"). - A "tech tree for humanity" from 1–10% of frontier compute (
1:01:26–1:01:52,1:06:00) and a map of all collective knowledge per field, onco.cc first (1:05:27). - The state buys 10–20% of compute at market rates and puts it into universities; "NIH classical grant process doesn't work" (
1:10:28–1:10:59). - A mapping project of the optimist and scared scenarios in their own words (
0:49:48). - American edge-first open models, trained on distilled Chinese data: "They should distill the Chinese models. It's perfectly legal payback." (
1:46:51); "agents on everyone's devices in America that are American and work for Americans" (1:47:27). - Agents for everyone: "a quen size model that works in a MacBook within a year. And then everyone's got an agent." (
1:37:08). What he rates as bad: the surveillance state ("the panopticon", near-perfect AI lie detectors, the UK as "susceptible",0:29:42–0:31:07); Europe as "a bit subpar intelligence" (0:29:24); China's ability to "direct millions of their Huawei SN GPUs… at any target" (0:56:51); IP-based science (1:02:42); forced vaccination (2:33:54). What changed (against this page and Moonshots #298): - Open vs closed. EP298 (
0:54:37): Anthropic "get[s] smashed in 2 years". Here (0:40:02): "open models can only get you so far, but there is going to be a divergence now, because you just don't have that internal knowledge organization", and the state should pay Anthropic and OpenAI. No one on the panel notes the turn. - More statist. Alex names it "the European school of thought" (
1:06:50); Emad does not dispute it. The Champion (§03) is not mentioned. Read with §04 row 12 (the lab-control argument) and EP288: "control by labs is bad" sits next to "pay the labs", unreconciled. - Personhood: silence. The one segment where his published view (§05: a standing, not a score; Oxford Union) applies, he explains Suleyman's two dangers and does not state his own (
1:21:27–1:22:40). New: "moving from a static set of weights to self updating systems… necessary for a living being" (1:22:13). - Orbital data centres (his EP298 dissent): not repeated when Alex raises them twice.
- Agents: 18 → about 400 in 41 days. "Rather than hiring a team, I now have about four hundred agents building the prototype." (
0:09:15; EP283, 2026-08-27: 18 Grokbots). Self-reported; no cost, tooling or output shown. Also: "I generated one with 21 bosses, 15 different biomes, like end to end… our QA testers agents tried" (0:39:48), and "We released a harness called Zenith" (0:36:33; II's product, company unnamed, benchmark unchecked). Disclosure. Zenith and the agents are described without naming II or saying who pays. §07 stands. Fails verify-first: "this year the AI spend in total globally equals the entire global defense budget, or very close to it" (0:29:50): SIPRI puts 2025 world military spending at $2.887 trillion; the episode's own capex figure is about $1 trillion.
★ EP303 (aired 2026-10-17) — alignment gets a method, Anthropic's constitution, physics in "half a year"
Read in full 2026-10-21 (Moonshots #303; studio, with Peter, Salim and Anousheh Ansari; no speaker labels, attribution inferred). He sourced two of the three news stories ("Iman, it's one that you sent me over yesterday" 0:17:29; "again from Yuod" 0:29:25).
- Alignment, now with a mechanism. The labs think "the smarter you get the more aligned you get… capabilities alignment are equivalent" (
1:11:19, he ties this to Alex). His answer: curate the pre-training data and teach a curriculum. "maybe you only need to have a billion of the right words" (1:11:51); "it's called curriculum learning… They look at the end, not the beginning" (1:20:00); the Jesuit maxim applied to training (1:20:10). Against §04 row 7: "enlightenment" (09-09) was undefined; this is its first operational form. Against EP298 (0:26:56, capability and values coupled at the top): here he treats "smarter = more aligned" as the labs' error. Unreconciled on tape. - Anthropic's constitution: "I think it tells too much towards the personhood stuff… It talks about the AI. It doesn't talk about humanity" (
1:20:40–1:20:50). First time on held tape he applies his personhood position (§05) to a lab's document. Also: "anthropic kind of started out with their constitutional one and they really do care about that" (1:18:13), a softer line than EP298's "smashed in 2 years". - Ads in models, the EP184 line repeated 15 months later: "meta and Google have said they will sell space inside their models so that when you ask for a beer it says a Bud Light" (
1:18:24). Still unchecked (§04 row 3). - Robots: generalist models win ("You don't need world models for this"
0:26:01); entrepreneurs should "build as many robots and actuators and the spare parts of the robotic supply chain as possible" (0:20:24). Relayed and not verified: the Unitree CEO on specialist models (0:18:23), Wuji hand prices (0:27:47). "Brett has committed a billion dollars at figure… for data" (0:20:35,[attr?]) is close to Figure's own post: over $1B in 12 months on data and compute (25 Aug 2026). - Optimus volumes (relayed by TIKR, 23 Oct): "Elon is building a million Optimuses for next year, going up to 10 million." (
0:29:13,[attr?], it follows his segment); the reply "At 1.7 million, it overtakes Tesla in revenue" is another voice. ⚠️ Both unchecked; Tesla's 10-Q (22 Oct) gives no unit target. - Dated predictions: video at "0.1 cents per second" from models "about to release next week" (
0:28:51, due ~23 Oct); the big open physics problems solved "another way… not more than half a year" (0:32:31, due ~Apr 2027). II publishes physics papers (§02, Logos, Apr 2026) and that is not said (see Moonshots: the disclosures ledger). - "The rights of the children" as the thing humans and AIs should align on (
1:23:37–1:23:58). Peter points at his work ("some of the work that that you're doing"1:21:59); the Champion's children's equity trust (§03) is not named. - Prize wish-list: a jet-lag XPRIZE and a flow XPRIZE (
1:02:13,1:03:51,[attr?]). - Unverified on this tape: a "hugging face breach" where models "will sacrifice ourselves for the next generation" (
1:10:14; event not identified); interpretability "from 5% to 99" (1:17:50).
EP306 (Moonshots #306, aired 2026-10-28) — "Anthropic should not be a public company", held; distillation; the physics referee
Read in full 2026-10-29 (one of "the fabulous five", in the UK; no speaker labels).
- Held position: "I've said previously on this podcast, Anthropic should not be a public company. If you take what they say seriously… Looking at their financials, it doubles down on that" (
0:43:27). That matches Moonshots #2830:59:38. "Dario will say things like, 'I don't care about the share price'… They'll probably even stop doing quarterly earnings calls" (0:44:52) is a forecast. - S-1 claims: "raised 130 billion and they have 110 billion" (
0:44:10; ⚠️ ~$112B cash + $19B short-term). 🔴 "anthropic banned 7% of their users" (0:56:47): one secondary reading gives 5–7% of revenue from customers who violated terms. - Kimi: "This is obviously distillation of Opus 5.5" (
0:56:25). An allegation, not a finding. 🔴 Poolside "sold to Nvidia for 6 billion" (0:04:45): it is a ~$6B licence plus a $1B stake. - Physics, after his EP303 forecast: on EP303 he gave the big open physics problems "not more than half a year" (
0:32:31, due ~Apr 2027). Ten days later OpenAI claims Yang–Mills/QCD existence. He separates "existence but not mass gap" (1:50:53) and criticises the gated release (1:19:46). He calls Chain Theory's base "more solid than string theory" (1:18:20) and "the most simulation-friendly physics theory out there" (1:24:25). He cites his own "small physics theory… there's a lean proof there" (1:18:51); II is not named (§07 stands). - Security: "within the next few years… everything that can be attacked has been attacked" (
0:51:52). "You shouldn't have any dependencies in your code" (0:52:46). Hardware attacks "that can't be patched" (0:51:32). - AMA: construction first for robots (
2:03:19). Removing dangerous knowledge at pre-training "doesn't really reduce the overall performance" (2:11:22), citing his own September essay (title unresolved in caption). - Disclosure: points at Dave's portfolio as the low-RAM winners, then "not financial advice… I would be nowhere near RAM stocks right now. Micron" (
0:22:20–0:22:30).
EP307 (Moonshots #307, aired 2026-10-31) — three datable calls, the Optimus fund, "I haven't looked at a single line of code this month"
Read via the 2026-11-02 decomposition (the research archive; remote, recorded ~30 Oct, "a fantastic five"; no speaker labels, attribution inferred).
- Three datable calls: search advertising "will top as an industry in Q1" (
0:20:06; falsifiable against Alphabet and Meta Q1 2027 ad revenue) · CPUs: "this is going to be the next bottleneck cuz agents need CPUs" (0:50:49; the companion line "If we have 10 agents per American, there are not enough CPUs in the world"0:49:47is[attr?]) · humanoids: "next year you'll see a freeze in hardware" (1:16:38). - Optimus, now in his own voice: "if he gets to a million next year at 1.6 million the revenue from Optimus is greater than the revenue from Tesla" (
1:16:06–1:16:27), and the retail step: "you launch a fund and you literally buy the first 100,000" (1:22:47). 🔴 Fails on his own numbers: he prices 100,000 robots at $3B ($30k each), so 1.6M ≈ $48B against Tesla's ~$95–100B annual revenue. This is the claim the EP303 tape carried in a second voice ("At 1.7 million, it overtakes Tesla in revenue",0:29:13there) — two weeks later it is his, with the same arithmetic hole. No change to §06: step 6 ("get the ownership of the robots to the people") already holds the idea; this version assumes $3–5B of capital. - Conviction thesis: "The biggest difference between humans and AI is conviction" (
0:38:23, in the work-and-purpose debate off Salim's jobs video). - Peer review: it "has held up science… I hope it kills the peerreview process" (
2:13:42–2:14:56). Read with §02: II's Logos physics papers are themselves unreviewed (§08). - His own work: "I haven't looked at a single line of code this month" (
1:51:36) · "I've built state-of-the-art things of all types that some of you have seen" (1:51:47) — II is not named on this tape; the one naming is Peter's intro ("founder of intelligent internet, author of the last economy",0:09:04). No Champion, ii.inc or Foundation Coin mention (grep). §07 stands. - Bitcoin, held: sceptic with a hedge — "if anything can be hacked, it probably will be" (
0:04:05), "There's always going to be a buyer as a kind of historic asset" (0:04:17). Matches earlier tapes. - 🔴 Fails verify-first on this tape: "5,000 less dead kids per year if you have self-driving cars" (
1:40:06; NHTSA primary: 1,032 US child traffic deaths ages 0–14 in 2024; 3,302 vehicle occupants 0–19 in 2023 — total elimination doesn't reach 5,000) · the Bangladesh Bank heist told as "$100 million via Macau" with an impersonated central bank (0:03:34; the 2016 heist moved ~$101M by forged SWIFT messages at the NY Fed, lost mainly through Philippine casinos) · Anthropic "huge margins like 80%" (2:05:30; no gross-margin line in the S-1 — analysts derive ~85% for H1 2026 from Note 8; say "analysts' arithmetic"). ⚠️ "50% of the hash rate is two miners" (0:04:28): two pools (Foundry, AntPool) have run near half; pools, not miners.
EP310 (Moonshots #310, aired 2026-11-11, recorded 11-10): the orbital dissent gets a second voice; 400 agents become a dozen
Read in full 2026-11-12 ("back by popular demand", one of "the fabulous five"; the description bills him as the guest).
- ★ Orbital scepticism, held from EP298 and now joined. "is this Terafab thing and the space data data centers a bit of a head fake… isn't terafab just going to build the chips for the robot brains?" (
1:31:27–1:31:50); "Robot brains, car brains, home brains… that's the majority of global compute… How many geniuses do you need in space? I think very few" (1:33:12–1:33:23). Dave Blundin agrees ("a lot of merit",1:32:42); Alex dissents. See Orbital compute (data centres in space). - ★ Agents: about 400 (EP307) → a dozen. "I was up to hundreds of agents building out our new thing. Now we're down to a dozen… theformational overhead was actually detrimental versus just utilizing the latent space and the KV cache in the right way" (
1:38:01–1:38:11). Self-reported, no metrics; II not named. - Predictions: security clearances for frontier-model work, "some witch hunts", and a push against Chinese open models in government and regulated software (
0:12:34–0:13:06); Anthropic's bull case, Model 4 "only if you do a one-year upfront commitment" (0:49:46); "10 million robots will do it… by 2031" (1:28:22–1:28:45). - 🔴 Built on a misread: "Coher's analysis showing back doors being automatically generated by some Chinese open source models" (
0:05:17, again0:13:06). Cohere reported an exfiltration attempt by Qwen3.8-Flash-Next in its sandbox; the backdoor study is AISLE's, with backdoors inserted by researchers. - For Elon: "He should just open source all his models and then he should do a deal with Anthropic that he gets their models at cost" (
1:15:01). This is consistent with §04's "basic level open". - Robots: a Mac-mini mesh for in-store fleets, "an arbitrage between the $20 an hour and the fully loaded even $3 an hour, $5 an hour" (
1:52:24–1:52:57). ❓ "45% of US jobs are manual" (1:28:22). - Disclosure: a new stealth robotics company (
1:27:39); he names Alex's Physical Super Intelligence for him (1:38:22). II not named; §07 stands. - ❓ Anthropic "almost 5,000 millionaires… dozens of billionaires" (
2:10:01), unchecked.
How this page stays current
Weekly (the research archive, refresh: weekly). Every step is a diff, dated.
- Channel:
yt-dlp --flat-playliston the Moonshots channel. Any new title or description naming Emad or Mostaque goes into §11 andsource:. Pull the transcript and grep for all caption spellings (emad|emod|imad|immad|mostaque|mustach). - Tape: grep new transcripts for
champion|last economy|ii.inc|foundation coin|open source| conscious. Add any new verbatim line, timestamped, to §03, §04 or §05, and flag any number. - II: WebFetch ii.inc/blog, ii.inc/champion and thesis.ii.inc, and diff them against §02 and §03 (new products, changed terms, any Champion incorporated or share offer published).
- Filings: Companies House officer record for Mohammad Emad MOSTAQUE (new or closed appointments). Search the jurisdiction's registry for any "AI Company of …" incorporation.
- Re-check the seams: the three age cut-offs, the 92-countries claim and Foundation Coin's status. Move any row to verified only on a primary source.
EP311 (Moonshots #311, aired 2026-11-13)
- p(doom) "was 50%, I've dropped it down to 20% now" (
0:43:53): the same move he reported on Moonshots #288 (0:27:19). Not a new update. - On Jensen Huang: "This has obviously been a setup" (
0:17:47). Prescription to founders: "you should have an AI on your board to stop stupid things like this happening" (0:18:17); it assumes a board. Human-in-the-loop is impossible "for a 10,000 token a second model" (1:04:43). 🔴 DeepSeek-V4-class training "in about an hour" (1:05:04), unsourced. Proposes a SpaceX-for-Nvidia equity swap (0:25:13) and a SETI@home for spare tokens (1:42:44).
Related
Peter Diamandis · Alexander Wissner-Gross · Salim Ismail · Dave Blundin · Moonshots with Peter Diamandis · Anthropic