Astro Teller

Who he is

Affiliation (role check run 2026-10-03): Captain of Moonshots and CEO, X, the moonshot factory, since 2010 — sixteen years, the longest continuous tenure of anyone on the Moonshots LIVE bill. Structural correction: X sat under Alphabet from 2015 to 2022 and has reported inside Google again since 2022 (Wikipedia); "Alphabet's moonshot factory" is shorthand, not the org chart. Board seats: Wing and Intrinsic (astroteller.net). Instructor (one of six) on the paid MasterClass certificate course Breakthrough Innovation Strategy with X (from $299; 2025-10-23). Host, The Moonshot Podcast (x.company). Born Eric “Astro” Teller, 1970, Cambridge, England; grandson of nuclear physicist Edward Teller. BS Computer Science and MS Symbolic Computation, Stanford; PhD in Artificial Intelligence, Carnegie Mellon, 1998 (advisor Manuela Veloso; Hertz Fellow). Former Stanford faculty. Co-founder and chairman of BodyMedia (wearables) and co-founder of Cerebellum Capital. Novelist — Exegesis (1997), Among These Savage Thoughts (2006), Golem — and co-author of Sacred Cows (2014) with Danielle Teller.
One-line position: Radical innovation is not a talent problem but a process problem — build an institution that makes killing your own best ideas cheap, fast and socially rewarded, and breakthroughs become a throughput question rather than a matter of genius.

Discipline & technical bet

A credentialed AI researcher who stopped doing research to build the machine that does it. The wager is portfolio-and-process over insight: run 100+ candidate moonshots a year (~2,000 evaluated over fifteen years, per his DLD/WIRED conversation, 2025-02-28), spend almost nothing per candidate, attack each one at its weakest point first, and accept a ~2% hit rate as the design target rather than a failure. The bet changed shape in the window and the September build missed it. In 2016 his product was “producing new Alphabet entities.” Since a January 2024 memo (“spinning out more projects as independent companies funded through market-based capital,” Bloomberg, 2024-01-22) and the arrival of Series X Capital — an independent $500M+ fund run by Gideon Yu, legally bound to invest only in X spinouts, with Alphabet as a minority LP (TechCrunch, 2025-11-02) — the output is companies that land “just outside the Alphabet membrane.” His own August 2026 post states the new goal plainly: evaluate hundreds of ideas a year to “produce a few truly world-changing moonshots each decade” (X blog, 2026-08-03). The technical bet underneath is unchanged and visible in what X now ships: AI as a component inside a domain problem — grid planning (Tapestry), disaster forecasting (Bellwether), plastics sorting by hyperspectral imaging (Materra), building-approval orchestration (Anori), free-space optics (Taara). He is still the only person on the Moonshots bill who runs an idea-to-kill pipeline at institutional scale and publishes its metrics — now with an outside fund marking the output to market.

Key claims (Says)

• Definitional: a moonshot needs three things — a huge real-world problem, a product or service that would eliminate it, and a breakthrough technology that offers hope of getting there. Held — restated verbatim at Disrupt (TechCrunch, 2025-10-27) and in his own post (2026-08-03).
• Process: “If it sounds reasonable, we're not interested” — reasonableness is a disqualifier by definition. Held — “if someone is proposing a moonshot and it sounds reasonable, the company isn't interested, because that, by definition, wouldn't be a moonshot” (2025-11-02).
• Process: attack the hardest part first and try to kill the idea cheaply. Ideas that test more crazy get terminated; ideas that test less crazy get incremental funding. Held — “I'm not giving you this money so you can succeed … I want to stop the 99% that are wrong as fast as possible” (DLD, 2025-02-28); “if it's a little bit more crazy than we thought, cool, high five, let's put a bullet in its head and move on” (2025-10-27). New in the window: he frames spinouts as a bias-removal device — “if you feel like 'this is my baby,' what are the chances I get you to practice real intellectual honesty?” — so teams get no equity until a project is ready to leave (2025-11-02).
• Empirical (self-reported): ~2% hit rate; 100+ projects started annually; 44% of spending goes to the graduates that become “outrageously good”; 5–6 years to exit for successful projects. Held as restated (2025-10-27), still unaudited and self-defined. Note the frame has softened in his own writing: “hundreds of ideas annually” and “a few truly world-changing moonshots each decade” (2026-08-03) — a decade-scale promise, not a 5–6 year one.
• Cultural: breakthroughs need equal parts audacity and humility — and creativity is relearnable in an environment where failure is genuinely safe. Unfalsifiable as stated — rewrite or retire. No observable exists; it is now the thesis of a $299 certificate course. Proposed observable: none that is honest. Keep as a value, not a claim.
• On AI: AI is “a component technology” — it makes systems smarter rather than replacing them. Held by revealed behavior — every 2025–26 X output embeds AI in a vertical problem (Tapestry, Bellwether, Materra, Anori, Heritable Agriculture); X has launched no model, no agent platform, no AI-as-product bet. Source for the view: IEEE Spectrum, 2016-12-08; for the behavior: project list, Anori, 2026-03-19.
• Contrarian, on robotics: “personal robotics” is not a moonshot because it is not a problem — it's a form factor (“'My laundry is on the floor, I wished someone else would pick up my laundry,' that's a problem,” IEEE Spectrum, 2016). Held (live) — see prediction below. Revealed behavior agrees: X's two robotics graduates were task/industrial (Everyday Robots, folded into DeepMind January 2023; Intrinsic, folded into Google February 2026 to be “the Android of robotics,” CNBC, 2026-02-28). He sits on Intrinsic's board. This still puts him against the humanoid wave most of the 2026 bill is long on.
• On Glass: a learning platform mistaken for a consumer product. Held — Glass Enterprise Edition ended 2023-03-15 (Wikipedia); nothing in the window changes the framing.
• On X's output: “producing new Alphabet entities” (2016). Drifted — the output is now companies outside Alphabet, by design: “landing it just outside the Alphabet membrane, where we can be very tight with them … but not necessarily control them, makes sense” (2025-11-02). Even Verily Health moved to the Series X structure with Alphabet at a “significant minority stake” (2026-08-03). The mechanism (output = companies) held; the destination changed, and the change is the story of the window.

Notable predictions — with falsifiable checks

• (ongoing) The moonshot-factory model is repeatable. Check: name X graduates after Waymo with independent enterprise value above $1B. Live — unconfirmed, and the post-2016 ledger leans against it. Post-Waymo graduates and their fates: Loon (shut 2021-01), Makani (shut 2020-02), Chronicle (absorbed into Google Cloud 2019), Mineral (independent 2023, shut 2024-07, tech licensed to Driscoll's — SiliconANGLE), Everyday Robots (folded 2023-01), Intrinsic (folded into Google 2026-02), Malta, Dandelion, Skip, Tidal, and the Series X cohort — Taara (2025-03), Chorus (2025-04), Heritable Agriculture (2025), Anori (2026-03, $26M Series A). None has a public valuation near $1B. Rewritten check: by 2028-12-31, does any Series X-backed spinout price a round at ≥$1B? Yes → Held; no, and Series X does not raise a Fund II → the model reduces to one extraordinary draw plus a well-told process story.
• (2016–) Ambient AI in ordinary devices — the coffeemaker that learns your order (IEEE Spectrum, 2016). Held, now unremarkable; the component-technology framing aged better than the humanoid one.
• (ongoing) Anti-humanoid position. Check: by 2028-12-31, a general-purpose humanoid deployed commercially at >10,000 units doing varied unstructured work. Held (live) — as of mid-2026 no humanoid from any maker is above the low hundreds of units in sustained commercial use; Figure has 40 units billing BMW ~$25/robot-hour (January 2026); Agility reports 65,000 operating hours across nine customer sites; Tesla's ~1,000–1,200 Optimus units are internal, with “zero external sales” (Technology.org, 2026-07-18; IIoT World). Two years of runway left on the check; the bill's humanoid optimists are the ones under pressure so far.
• (ongoing) Near-zero delivery cost would sharply increase resource sharing. Check: a published per-delivery cost curve from a drone operator at scale. Live — still untested. Wing is now Walmart's drone partner in four metros, 66 stores, contracted to 270+ locations by 2027, and declined to disclose its per-delivery cost when asked by multiple outlets; its predecessor DroneUp spent ~$30 per delivery against a $7 target and shut down (TechTimes, 2026-07-30). Customer price is $19.99 or free with Walmart+ — a price, not a cost.
• Structural claim: the failure database lets X revive killed ideas. Check: a named revived-and-shipped project. Held, with a nuance. Taara is the laser-link technology built inside Loon; Loon was shut in January 2021 and Taara spun out as an independent company in March 2025 with Series X backing, now live in a dozen-plus countries including a Congo River link and Rio favelas, and shipped a 25 Gbps optical-phased-array product at MWC in February 2026 (Taara; SiliconANGLE, 2026-02-23). Nuance: Taara was salvaged from a dying project, not resurrected from the database years later — a kill that saved its best organ, which is a stronger story than the one he usually tells and a weaker test of the database.

Revealed behavior (Does)

• Kills things and says so — the record extends. Makani (airborne wind) closed February 2020 — “the road to commercialization is longer and riskier than hoped” (Forbes, 2020-02-18). Loon closed January 2021 — Teller: “the road to commercial viability has proven much longer and riskier than hoped” (CNBC, 2021-01-21). Mineral closed July 2024 after a year as an independent Alphabet company. Glass ended March 2023. Dates now verified against dated press, closing the September caveat. Same sentence, twice, four years apart — the kill script is literally a script.
• Took the lab outside the balance sheet. January 2024: dozens of support-staff layoffs and a memo pivoting to “market-based capital” under Ruth Porat's cost discipline (Bloomberg, 2024-01-22). 2025: Series X Capital, $500M+, Alphabet a minority LP, “legally obligated to invest exclusively in companies spinning out of X” (TechCrunch, 2025-11-02). The man who preached that the method is general finally had to run it with someone else's money.
• Four spinouts in eighteen months — Taara (2025-03), Chorus (2025-04), Heritable Agriculture (2025), Anori (2026-03, $26M led by Prologis with Series X) — plus Verily Health moved into the structure (X blog, 2026-08-03). Compare: the prior four years produced mostly shutdowns and re-absorptions.
• Maintains an institutional memory of failure — “we don't want someone who gets hired two years after we kill a project to come up with the same idea and then spend three months rediscovering it” (IEEE Spectrum, 2016). This is _dropped.html as corporate infrastructure.
• Spends asymmetrically. 44% of budget on the winners means the other 56% buys information about what not to build (2025-10-27).
• Stayed sixteen years in the same seat through two reorganizations (into Alphabet 2015, back into Google 2022) and a funding-model inversion. Not optimizing for personal narrative velocity.
• Teaches the method — now for money. The Moonshot Podcast (ten-part series from SXSW 2025), the MasterClass certificate course with X (2025-10-23), the Moonshots LIVE masterclass (2026-09-25, moonshots.com). The process is deliberately not a trade secret; it is now also a product line.

Feels

Comfortable with loss in a way that appears trained rather than temperamental — the emotional work of the job is making other people's grief over a killed project survivable, and he has built ritual around it. Reads as playful rather than evangelical. New in the window: a visible relief at having found a way to keep moonshots alive without asking Alphabet to own them — the August post reads like a man who solved his own bottleneck.

Hears

Two audiences that pull opposite ways: Alphabet capital allocators (Porat's cost discipline since 2024) and a public that wants wonder. A third voice has joined: outside LPs — family offices and sovereign funds behind Series X — who will price his graduates. He has kept all three by making failures part of the story. What he still hears least is the outside view on whether the method works at n=1 with no balance sheet at all.

Sees

Sees process where others see genius, and sees form factors as a category error. His characteristic move is to find the load-bearing assumption and test it first. The blind spot the September build named — scale-dependence — is now partly tested from the inside: he had to invent a $500M outside structure because Alphabet could “only accommodate so many Other Bets.” He saw the capital constraint; what he has not yet said is what the method looks like with none.

Incentive map

Re-derived this cycle — the September build's “unusually clean” needs a downgrade, though it stays the cleanest on the bench. Salaried Google executive; no fund of his own; no carry in Series X Capital found in any source (the fund is Yu's, independent, legally fenced to X spinouts — absence of evidence, flagged not asserted). Corrections: he does sell a course — one of six instructors on a $299 MasterClass certificate program built on X's method (2025-10-23) — and he holds board seats at Wing and Intrinsic. Structural stake: his job exists only if moonshots are believed to pay, and since 2024 the belief has an outside price — Series X's returns will grade him in a way Alphabet's Other Bets line never did. Cannot easily say: that the model needs patient capital at a scale no startup has; that the graduate record after Waymo is thin; that a 2% hit rate is indistinguishable from luck at his sample size; that the pivot to outside money was forced rather than chosen. Self-reported metrics remain self-reported. Moderate discount. Grade him on named spinouts, dated shutdowns, and Series X's portfolio marks — all public.

Theories aligned with

• Portfolio theory applied to R&D — many small bets, asymmetric payoff, losses as cost of information; now with an actual venture fund attached
• Fail-fast / lean experimentation, at capital scale rather than startup scale
• Techno-optimism — the engineering-discipline variant, not the abundance-rhetoric variant; the least evangelical optimist on the bill
• Anti-humanoid / task-specific robotics — a live minority position, holding as of mid-2026
• Institutional memory as infrastructure — the failure database as a first-class asset
• Venture-studio model — X now resembles Ng's AI Fund in structure (co-found, then fund), with a longer fuse and a captive LP

What he's reacting against

• Innovation as a talent search — breakthroughs from hiring geniuses rather than building a process
• Reasonable proposals — explicitly disqualifying; incrementalism dressed as ambition
• Form-factor thinking — “personal robotics,” “the metaverse,” any category defined by shape rather than by a problem
• Sunk-cost loyalty — and, new in the window, founder ownership bias: no equity until the project leaves, so nobody has a baby to protect (2025-11-02)
• Failure shame — the emotional cost of killing work as the actual bottleneck
• NEW: the corporate membrane itself — Alphabet as sole owner became the constraint on throughput; the fix was to make the membrane permeable

Where he overlaps / splits (with Rich)

• The strongest operational overlap on the entire Moonshots bill. Your Forge runs a promote/kill sweep every 7th run with _decisions as the append-only system of record; your Ideas folder has graduation thresholds; your _dropped.html is a managed reject list. Teller runs precisely this system with a $100M+ budget, a decade of data, and — since 2025 — an outside fund grading the graduates.
• The questions worth asking him are yours already: what fraction of kills get revived and ship (Taara is now his one named answer — press on whether it counts)? What is the honest false-negative rate? How do you keep a kill sweep from becoming a formality once the team knows the ritual? New: did spinning out change what you kill — do outside LPs make you keep things you would have shot?
• A real split worth sitting with — §12 Embrace the Tensions. Teller says try to kill your best ideas, hardest part first. You hold that the “I wonder if” is where the fun lives (closest canon: §19 Dream & Ask What-If; §16 Would I Do This For Free?), and that the moment it becomes “I must prove,” the fun dies. Not compatible instincts; the tension is load-bearing. His method is right for a portfolio with a budget and a board; yours is right for a one-person bench where the scarce resource is appetite, not capital. Killing your best idea on Tuesday is cheap for Alphabet and expensive for you. Worth naming rather than resolving. The September build cited “§22” and “§14” for this — neither exists as cited; the canon stops at §19 and §14 is Don't Make Me Think. Corrected.
• Overlaps on §05 Trust the Systems, §09 Judgment Over Execution, §10 Research Like a Scientist — “attack the hardest assumption first” is §10 with a budget attached. (September build cited §5/§10/§13 for these; renumbered against the canon file.)
• Splits with Diamandis on register — same optimism, no urgency rhetoric, no timeline claims. Useful as a control on the rest of the bill, though he now has a $299 course too.
• Splits with the humanoid-robotics consensus in parts of Blundin's and Wood's theses — and, as of mid-2026, the deployment data is on his side.
• Overlaps with Ng on structure (venture studio; AI as component in applications) — neither cites the other.
• For the bench: he fills a genuine hole — operators who run selection systems. The window added the piece that was missing: what happens to a selection system when its patron's patience runs out.

Track record

Graduates (into Alphabet/Google): Waymo (2016; the clearest commercial success in autonomy), Wing (2018; now Walmart's drone partner, 66 stores, contracted to 270+ by 2027), Verily (2015; moved to the Series X structure 2026), Google Brain (into Google 2012), Chronicle (into Google Cloud 2019), Intrinsic (2021; into Google 2026). Spinouts (outside, Series X era): Taara, Chorus, Heritable Agriculture (2025), Anori (2026). Earlier independents: Malta, Dandelion, Skip, iyO. Shutdowns: Makani (2020-02), Loon (2021-01), Everyday Robots (2023-01), Glass (2023-03), Mineral (2024-07). Tenure: sixteen years, same lab, three org charts. Other Bets context: Q2 2026 revenue $382M, operating loss $1.8B, versus $373M / $1.24B a year earlier (9to5Google, 2026-07-22) — the segment is Waymo-dominated and the loss is widening, which is the pressure behind the pivot.

The honest read: a remarkable record and a front-loaded one. Waymo, Wing, Verily and Brain all originate in X's first half. The years 2019–2024 produced more funerals than weddings. The strongest case for him is that he closed Makani, Loon and Mineral anyway, and then re-engineered the lab's funding when the patron tightened. The strongest case against is that a 2% hit rate over ~15 years is a small enough sample that one Waymo carries the entire result — and Series X's marks will be the first outside audit of that.

Empirical vs normative

Empirical (testable, self-reported): 2% hit rate, 100+ projects/year, ~2,000 evaluated, 44% of spend on graduates, 5–6 year exit. Unaudited and internally defined — treat as directional. Empirical (externally verifiable): the graduate, spinout and shutdown list — now with dated press for every closure — and, from here on, Series X's portfolio. Normative: that reasonable ideas are not worth doing; that failure should be socially safe; that form-factor framing is a mistake; that nobody should own their baby until it can walk. Cleanly separated in his own speech — he does not usually dress values as data.

Weak spots / open questions

• Sample size. 2% on a portfolio this small cannot be distinguished from luck. He presents it as a validated design parameter.
• Capital dependence — now partly answered, uncomfortably. The method did not generalize below Alphabet scale; it needed a $500M dedicated fund and a captive strategic LP to keep running. For your purposes the crux stands: does the kill discipline survive at n=1 with no balance sheet?
• Front-loaded graduates. No post-Waymo graduate near $1B. The Series X cohort is the test; Anori's $26M Series A is the biggest disclosed round.
• Self-defined metrics. “Graduate” now spans absorbed-into-Google, independent-Alphabet-company-then-closed, and Series X spinout — three very different outcomes under one word.
• False negatives are invisible. Taara is the one revival story and it is a salvage, not a resurrection. He has still never published a wrong-kill.
• The humanoid position is his live exposure — currently winning; two years of check remain.
• NEW: incentive creep. A paid course, a podcast, two board seats, and a fund whose returns now grade him. Still the cleanest on the bench; no longer “sells no course.”
• NEW: what did Moonshots LIVE add? He gave the masterclass on 2026-09-25; no transcript or dated write-up of its content was reachable this pass. If Rich attended, his notes beat anything here.

Rich's take

• (your synthesis here)

Delta log

2026-10-03 — validation pass (CSS to v2.5; template v2 kept)

• Grades: 9 Held / 2 Drifted / 1 Wrong (plus 2 live predictions with rewritten dated checks, 1 unfalsifiable claim marked for retirement). Backup: _archives/astro-teller.html.bak-20261003.
• Wrong: Incentive map “sells no course” — he is an instructor on a paid MasterClass certificate course with X, from $299 (2025-10-23), launched six weeks before the September build; also missed Wing and Intrinsic board seats (astroteller.net).
• Drifted: “producing new Alphabet entities” → companies outside the membrane (TechCrunch, 2025-11-02; X blog, 2026-08-03); “runs no fund / unusually clean” → Series X Capital ($500M+, Gideon Yu, Alphabet minority LP) now marks his output to market; no carry found, flagged not asserted.
• Held: three-part definition and “reasonable = no” (TechCrunch 2025-10-27, 2025-11-02); kill-first (DLD, 2025-02-28); self-reported metrics as restated; AI as component (project list; Anori, 2026-03-19); anti-humanoid (2026-07-18); Glass framing; ambient AI; revive-a-kill via Taara (Taara; 2026-02-23).
• Live: repeatability (no post-Waymo $1B graduate; check rewritten to Series X marks by 2028-12-31); near-zero delivery cost (Wing declines to disclose, 2026-07-30).
• Most surprising delta: X stopped making Alphabet companies. Layoffs and a market-capital memo (Bloomberg, 2024-01-22), then a $500M+ outside fund, then even Verily Health moved into it (2026-08-03). The September build, written five weeks after Series X was public, carried the 2016 model. Second surprise: the revival question Rich wanted to ask him has a named answer — Taara came out of Loon.
• Closed the September caveat: Makani 2020-02 (Forbes) and Loon 2021-01 (CNBC) verified; added Mineral 2024-07 (SiliconANGLE), Everyday Robots 2023-01, Intrinsic into Google 2026-02 (CNBC). Org-chart correction: X reports inside Google since 2022.
• Principle citations corrected against _LIFE-PRINCIPLES.html: v1's §22 and “§13 first principles” do not exist as cited; now §12, §19/§16, §05, §09, §10.
• Not reached this pass: Fast Company's 2026 X feature (Materra, Bellwether — paywalled), any write-up of the 2026-09-25 masterclass. Confidence stays medium until one of those lands; the structural picture is now sourced from X's own blog, so the floor is higher than in September.
• Tier: semiannual — the model moves on an annual cadence (Jan 2024, Nov 2025, Aug 2026) and his talk restates a 2016 script between moves; spinouts, shutdowns and Series X events are wake triggers rather than grounds for quarterly.

2026-09-05 — created

• New persona, built at v2 depth for Moonshots LIVE 2026-09-25, where he gives the masterclass on building moonshots.
• Verified this pass: biography, degrees and dates (Wikipedia); the three-part moonshot definition, kill process, 2% hit rate, 100+/year, 44% spend, 5–6 year exit, Waymo/Wing (TechCrunch, 2025-10-27); component-technology AI view, anti-humanoid position, failure database, Glass framing (IEEE Spectrum).
• Caveat logged: the IEEE Spectrum interview is dated — it discusses Loon and Makani as active and self-driving as “currently graduating,” placing it around 2016. Its views are used here; its project status is not. Makani (2020) and Loon (2021) shutdown dates are carried from general knowledge and were not re-verified against primary sources this pass. Verify before quoting the dates to anyone. (Closed 2026-10-03 — see above.)
• Framed the two open questions that matter to Rich: post-Waymo graduate record, and whether the kill discipline survives without a balance sheet.
• Named the tension explicitly rather than smoothing it — kill-your-best-idea vs protect-the-wonder is a genuine conflict, not a misunderstanding.
• Confidence: medium. Raise after checking the post-2018 graduate list against a primary source.

Sources

• Primary — X and Teller: X blog — “Bringing Moonshots to Market” (Teller, 2026-08-03) · The Moonshot Podcast · astroteller.net — About · Wikipedia — Astro Teller
• Interviews, 2025: TechCrunch Disrupt (2025-10-27) · TechCrunch — spinouts and Series X (2025-11-02) · DLD Munich with WIRED (2025-02-28) · NPR (2025-09-12) · IEEE Spectrum (2016-12-08; views only)
• Structure and funding: Bloomberg via Yahoo — layoffs, outside capital (2024-01-22) · Wikipedia — X Development (project table, org history) · 9to5Google — Alphabet Q2 2026 Other Bets · MasterClass × X certificate course (2025-10-23)
• Spinouts: Taara — spin-off and Loon origin (2025-03) · Wikipedia — Taara · SiliconANGLE — Taara Beam (2026-02-23) · TechCrunch — Anori (2026-03-19)
• Shutdowns and absorptions: Forbes — Makani (2020-02-18) · CNBC — Loon (2021-01-21) · SiliconANGLE — Mineral (2024-07-03) · CNBC — Intrinsic into Google (2026-02-25) · CNBC — “Android of robotics” (2026-02-28)
• Prediction checks: Technology.org — humanoid deployments (2026-07-18) · IIoT World — Optimus 2026 · TechTimes — Wing Orlando, DroneUp cost (2026-07-30) · Talk Business — Walmart–Wing expansion (2026-01)
• Event: Moonshots LIVE 2026 (2026-09-25) — masterclass
• Not reached: Fast Company, “Google's Moonshot Factory is just getting started” (2026; paywalled — Materra, Bellwether); any dated write-up of the 2026-09-25 masterclass. Next pass.

validated 2026-10-03 (template v2 kept; CSS v2→v2.5) · backup: _archives/astro-teller.html.bak-20261003 · converts-from: astro-teller.html (2026-09-05 build) · Singularity 🧠 / People