EVIDENCE FILE
Evidence File - The Robots We Count and the Robots We're Promised
Every claim in the article resolves to a numbered source record below or to an explicitly declared gap. Sources were retrieved and read; where a source could not be retrieved, the gap is recorded rather than filled.
0. HOW TO READ THIS FILE
Every claim in the companion article is supported by a numbered source record here; the draft cites by document rather than by tag. The tags are:
| Tag | Meaning |
|---|---|
| [FACT] | Established by a primary source I retrieved myself |
| [CORP] | Corporate statement — a company or its executive asserting something about their own product or plans |
| [FCST] | Forecast — a prediction about the future, attributed to whoever made it |
| [SEC] | Reported by secondary press; I did not retrieve the underlying primary record |
| [GAP] | Not retrieved. Named as a gap. Nothing is claimed |
Rule applied throughout: a number I could not retrieve is a [GAP], not an estimate. Where the brief supplied a source I could not open, I say so and do not paraphrase what it probably says.
1. SOURCE RECORDS
S1 — Google DeepMind, From AGI to ASI — [FACT] RETRIEVED
- Locator: arXiv:2606.12683 [cs.AI] — https://arxiv.org/abs/2606.12683
- DOI: 10.48550/arXiv.2606.12683
- Authors (14): Tim Genewein, Matija Franklin, Alexander Lerchner, Laurent Orseau, Samuel Albanie, Adam Bales, Cole Wyeth, Stephanie Chan, Iason Gabriel, Joel Z. Leibo, Allan Dafoe, Marcus Hutter, Thore Graepel, Shane Legg
- Dates: submitted 10 June 2026 (v1); revised 30 August 2026 (v2, "minor fixes and addition of related work")
- Retrieved: 2 Nov 2026, abstract page, full text
- Verbatim extracts:
- "Over the last decade, building human-level artificial general intelligence has moved from far-fetched speculation to being a concrete next-decade target for many of the largest AI organisations."
- ASI "can intuitively be understood as a system that is more intelligent and cognitively capable than large organisations of humans."
- Four pathways: "scaling AGI, AI paradigm shifts, recursive improvement, and ASI emerging from large-scale multi-agent collectives."
- "Due to large uncertainties for predicting ASI progress, it cannot be ruled out that AI progress might continue to accelerate over the next years."
- Load-bearing caution: "This could imply that the image of a single transformative step change, caused by the introduction of human-level AGI into our society, could be inaccurate. More apt might be the prospect of a series of transformative societal changes caused by AI-enabled progress and breakthroughs across many areas of science and technology."
- Why it matters: this is a frontier lab's own paper, and it argues against the very framing the headline implies. The strongest structural counterweight in the file is not a critic — it is DeepMind.
S2 — International Federation of Robotics, World Robotics 2026 — [FACT] RETRIEVED
- Locator: https://ifr.org/ifr-press-releases/news/five-million-robots-now-operate-in-factories-globally
- Dateline: Frankfurt, 24 September 2026
- Retrieved: 2 Nov 2026, full press release text
- Figures:
- Global operational stock of industrial robots rose 9% to a record 5 million units in 2025.
- Annual installations rose 11%, more than 600,000 new units in 2025.
- Jane Heffner, IFR President: "The new mark of five million robots operational in factories worldwide is more than double the number seven years ago."
- China: 354,000 units installed in 2025 (+20%), 59% of global deployments.
- United States surpassed Japan to become the second-largest market; installations +12% to almost 38,500 units.
- Forecast: installations +9% to 655,000 in 2026; 806,000 in 2029.
- SCOPE WARNING — mandatory for the draft: these are industrial robots, not humanoid robots. This release contains no humanoid unit count. Any sentence in the article that puts "5 million" next to "humanoid" is false and must not be written. Scope of this warning is the INDUSTRIAL release only. IFR's separate Service Robots release of 30 September 2026 does carry a humanoid figure — about 7,000 full-size units sold in 2025 — so no claim may be made that IFR counts no humanoids. See S13, added 4 November 2026 after an independent reviewer caught the article asserting exactly that.
S7 — IFR definition of "industrial robot" — [FACT] RETRIEVED
- Locator: https://ifr.org/industrial-robots
- Retrieved: 2 Nov 2026, full page text
- Verbatim: IFR's use of the term is based on ISO 8373:2021 — an "automatically controlled, reprogrammable multipurpose manipulator, programmable in three or more axes, which can be either fixed in place or fixed to a mobile platform for use in automation applications in an industrial environment."
- Why this source exists: an earlier draft of this article described the five million as "the bolted-down arms that weld, paint and palletise," implying no humanoid could be among them. That was wrong, and an independent reviewer caught it. The category is defined by use environment, not by form factor, so a factory humanoid falls inside it. The article's central contrast survives, but it must be stated as a contrast of place, not of shape.
- Also relevant: IFR's statistical department compiles installations of "multipurpose industrial robots" for around 40 countries, broken down by application, customer industry and robot type.
S3 — International AI Safety Report 2026 — [FACT, existence and scope only] PARTIALLY RETRIEVED
- Locator: https://internationalaisafetyreport.org/ and /publications
- Published: 3 February 2026
- Lead author: Yoshua Bengio (Turing Award laureate)
- Scale as stated by the publisher: "authored by over 100 AI experts," backed by "over 30 countries and international organisations"; described as "the largest global collaboration on AI safety to date."
- Structure as stated by the publisher: three central questions — "what can general-purpose AI do today, what emerging risks does it pose, and how can those risks be mitigated?"
- NOT RETRIEVED: the report body text and the Extended Summary for Policymakers. No finding from this report is quoted or paraphrased anywhere in the draft. Its role in the draft is limited to: it exists, when, who led it, and how many countries back it.
S4 — Musk at the World Economic Forum, Davos, January 2026 — [SEC]
- Secondary source retrieved: IBTimes UK, "'More Robots Than People': Elon Musk's Prediction at World Economic Forum, Explained," Tanmay Puri, published 23 January 2026 — https://www.ibtimes.co.uk/more-robots-people-elon-musks-prediction-world-economic-forum-explained-1773193
- Provenance note, 4 November 2026: the retrieved page carries the sub-heading "AFP News" above the report body, so the piece is AFP copy carried by IBTimes UK rather than original IBTimes reporting. The article now says exactly that. An earlier revision of this record omitted the AFP attribution, and a prior review consequently objected to the phrase "AFP-sourced reporting"; that objection is withdrawn — the phrase was accurate, and the omission was this file's.
- Setting: in conversation with BlackRock CEO Larry Fink; session titled "Conversation with Elon Musk | World Economic Forum Annual Meeting 2026."
- Quoted as reported: "By the end of next year I think we'd be selling humanoid robots to the public." / "My prediction is there'll be more robots than people... everyone on Earth is going to have one and going to want one... who wouldn't want a robot to... watch over your kids, take care of your pets... we are in the most interesting time in history."
- Also attributed: AI could become smarter than any individual by end of year, and exceed the collective intelligence of humanity within five years.
- Caveat: I did not retrieve a WEF transcript or video transcript. These are attributed quotations carried by press. The draft attributes them as reported remarks, never as a verbatim transcript I read.
- Useful context inside the same piece: IBTimes supplies its own comparison, quoted verbatim: "Industry forecasts, for example, estimate that by 2060 the world might have around 3 billion humanoid robots in operation, far fewer than the projected global population of over 10 billion in that year, meaning one robot for every three people rather than a robot majority." CORRECTION, 4 November 2026: this record previously described that forecast as "two orders of magnitude apart" from Musk's claim and placed it "inside 20 years". Both were wrong against this source. The source's own comparison is roughly one robot for every three people — less than half a single order of magnitude — and its date is 2060. The false characterisation is withdrawn, and the article now carries the source's own formulation instead.
S5 — Musk at a G20 virtual session, 1 September 2026 — [SEC]
- Secondary source retrieved: Breitbart, "Elon Musk's G20 Prediction: 1 Billion Humanoid Robots Will Outproduce All of Humanity in 10 Years," Lucas Nolan, 2 September 2026 — https://www.breitbart.com/tech/2026/09/02/elon-musk-predicts-1-billion-humanoid-robots-will-outproduce-all-of-humanity-in-10-years/
- Provenance note: this article is itself citing CNBC. So these are CNBC's account of Musk's remarks, relayed by Breitbart — second-hand twice. Treated accordingly: [SEC], never [FACT].
- Attributed claims:
- Within 10 years, 1 billion humanoid robots, with Musk calling his own forecast conservative and projecting "well over a billion."
- Each roughly five times as productive as a person; the fleet combined would outproduce all of humanity.
- Humanoid usefulness = three factors: AI software, onboard AI chip, electromechanical dexterity (especially hands).
- AI chip power shortfall of at least 15 gigawatts by 2027.
- Chip production climbing 40–50%/year against power generation outside China at 10–20%/year.
- $20–30 trillion/year, a 20–30% boost to the global economy, from digital AI alone.
- Software reaching "Stockfish-level" in about 12–18 months.
- Separately reported, same article: on the Moonshots podcast, Musk said of retirement savings: "Don't worry about squirreling money away for retirement in 10 or 20 years. It won't matter."
S6 — Tesla Optimus programme status — [SEC]/[CORP] via tertiary
- Source retrieved: Wikipedia, "Optimus (robot)" — https://en.wikipedia.org/wiki/Optimus_(robot) — which cites Tesla's Q1 2026 earnings call and prior reporting.
- Why tertiary: Tesla's own investor site refused access at the time of writing. That does not apply to the filings, which were retrieved on 4 November 2026 - see section 1c and GAP-1.
- What it records:
- Q1 2026 earnings call, 22 April 2026: Musk said the Gen 3 unveiling would be pushed "closer to production," citing competitors copying earlier designs; production expected to start around "late July, August."
- Tesla stated first-generation production lines "are being installed" at Fremont, "in anticipation of volume production," designed for 1 million robots a year; a second line at Gigafactory Texas designed for long-term annual production of 10 million robots, starting 2027.
- June 2024: Musk claimed limited production in 2025, with over 1,000 used in Tesla facilities and possible production for other companies in 2026.
- Stated eventual cost estimate: $20,000–$30,000 per robot.
- [CORP] vs [FACT] distinction: "designed for 1 million a year" describes a factory's rated capacity, not robots produced. No unit count for 2026 was retrieved. The draft must not convert capacity into output.
- Documented autonomy caveat — the strongest limitation evidence in the file:
- January 2024: viewers of a video of Optimus folding a shirt spotted a gloved hand at the edge of frame mirroring its movements, indicating teleoperation rather than autonomy.
- October 2024, "We, Robot" event: critics noted the robots were mainly teleoperated to interact with crowds; Tesla was criticised for not being transparent about this.
- Significance: the gap between a demonstration and verified autonomous capability is documented in this programme's own history. This is the single most useful fact for testing "verified autonomy" in the draft.
1b. SOURCES ADDED 3 NOVEMBER 2026 — THE MERGED SECTIONS
The four sections merged into the draft on 2 November ("The work we still tax", "The word we keep avoiding", "Who holds the switch") quote five sources that had never been entered in this file. They are recorded here, each with a locator, a retrieval date and a verbatim extract. Every extract below was re-retrieved and character-checked on 3 November 2026 by the lead, independently of the worker that first verified them. Where the two disagree, the discrepancy is stated.
S8 — Acemoglu & Restrepo, Robots and Jobs — [FACT] RETRIEVED
- Locator: https://www.nber.org/papers/w23285 — NBER Working Paper 23285, DOI 10.3386/w23285
- Issue date: March 2017
- Published version: Journal of Political Economy, 128(6), 2020, pp. 2188–2244 (per the NBER page's "Published Versions" field)
- Retrieved: 2 Nov 2026 (research pass) and again 3 Nov 2026 (lead re-retrieval, full abstract text, content check passed)
- Verbatim extract, the sentence the draft quotes:
"According to our estimates, one more robot per thousand workers reduces the employment to population ratio by about 0.18-0.34 percentage points and wages by 0.25-0.5 percent."
- Match: exact, character for character, including the hyphenated ranges.
- Scope, mandatory for the draft: the estimate concerns industrial robots in US local labour markets (commuting zones), 1990–2007. It is not evidence about humanoid robots and not evidence about generative AI. The abstract's own wording is "industrial robot usage between 1990 and 2007 on US local labor markets."
- Provenance caveat: the quoted sentence was read on the 2017 working-paper abstract. The JPE version is paywalled from this station and was not opened, so this file does not assert that the published wording is identical. The draft cites JPE as the peer-reviewed home and NBER as the text read — that is the honest formulation, and it is why both appear in the article's parenthetical.
- Why it matters: this is the strongest measured estimate in the package on automation and employment, and it is an estimate about a different technology. That tension is the section's point, not a weakness in it.
S9 — IMF Notes 2026/002, Global Economic and Financial Implications of Artificial Intelligence — [FACT] RETRIEVED
- Locator: https://www.elibrary.imf.org/view/journals/068/2026/002/article-A001-en.xml
- Full printed title: "Global Economic and Financial Implications of Artificial Intelligence: Lessons from a Scenario Planning Exercise"
- Publication date: 3 April 2026 · Volume 2026, Issue 002 · DOI 10.5089/9798229042482.068.A001
- Retrieved: 2 Nov 2026 (research pass) and again 3 Nov 2026 (lead re-retrieval; section heading "Fiscal Policy: Pressures on Revenue and Expenditure" located, string checks on the quoted text and on the qualification below both passed)
- Verbatim extract, the passage the draft quotes:
"Erosion of the labor tax base will necessitate adaptation of tax systems. Declines in employment and wage income weaken personal income tax bases, particularly under rapid diffusion, even after accounting for potentially higher wages for remaining workers. Shifting taxation from labor to capital through higher corporate income taxes may only partially offset these losses given capital mobility, market concentration, and international tax competition."
- Match: exact.
- THE QUALIFICATION THE DRAFT MUST CARRY. The quoted sentences end mid-paragraph. The same paragraph continues:
"…Therefore, tax systems should rebalance toward better capturing economic rents while minimizing distortions on investment decisions, supported by a stronger AI-enabled tax administration. Avoiding blunt "robot taxes," strengthening capital income taxation at the individual level with safeguards against base erosion, and leveraging AI to improve compliance and enforcement are important across scenarios."
- Verified independently in this entry's second retrieval: the strings "Avoiding blunt" and "robot taxes" both occur in the retrieved page text. So the source does not support an automation-levy reading; it explicitly advises against blunt robot taxes. Cutting the quotation at "international tax competition" is not a misquotation — the words are verbatim and the cut lands on a sentence boundary — but it suppresses the source's own opposite recommendation. The article must carry the qualification, and does.
- Classification, mandatory: this is scenario analysis, not an empirical estimate and not a forecast with a point estimate. The note synthesises a closed-door workshop held 10–11 December 2025 with the Economics of Transformative AI Initiative (EconTAI) at the University of Virginia, with roughly 50 participants, and states that "In both scenarios, transformative AI technologies were assumed to attain the capacity to perform a wide range of cognitive and manual tasks by the early 2030s." Attribute as IMF staff scenario analysis. Never as measured data.
- Access note: direct browser navigation to this URL hit a verification wall (HTTP 405) on the research pass; the body was returned by the fetch route. The second retrieval succeeded by the same route.
S10 — Butlin, Long et al., Consciousness in Artificial Intelligence — [FACT] RETRIEVED
- Locator: https://arxiv.org/abs/2308.08708 — arXiv:2308.08708v3 [cs.AI], DOI 10.48550/arXiv.2308.08708
- Title: "Consciousness in Artificial Intelligence: Insights from the Science of Consciousness"
- Authors: 19. Patrick Butlin, Robert Long, Eric Elmoznino, Yoshua Bengio, Jonathan Birch, Axel Constant, George Deane, Stephen M. Fleming, Chris Frith, Xu Ji, Ryota Kanai, Colin Klein, Grace Lindsay, Matthias Michel, Liad Mudrik, Megan A. K. Peters, Eric Schwitzgebel, Jonathan Simon, Rufin VanRullen. The page's own counter reads "by Patrick Butlin and 18 other authors."
- Dates: submitted 17 Aug 2023 (v1 00:10:16 UTC); v2 21 Aug 2023; v3 22 Aug 2023 — the version cited, correct.
- Retrieved: 2 Nov 2026 (research pass) and again 3 Nov 2026 (lead re-retrieval, full abstract, both quoted sentences character-checked)
- Verbatim extracts, the two sentences the draft quotes:
"From these theories we derive "indicator properties" of consciousness, elucidated in computational terms that allow us to assess AI systems for these properties." "Our analysis suggests that no current AI systems are conscious, but also suggests that there are no obvious technical barriers to building AI systems which satisfy these indicators."
- Match: exact on both, including the internal quotation marks around "indicator properties".
- Classification: this is an arXiv preprint, not a peer-reviewed journal article, and the draft cites it as an arXiv preprint. Not a journal citation.
- Why it matters: the second sentence is the strongest single support in the package for the article's intelligence-versus-consciousness distinction. It asserts both halves — nothing today is conscious, and nothing technical blocks building something that meets the indicators — and quoting it whole does the work without inference being added.
S11 — Regulation (EU) 2024/1689, Article 55 — [FACT] RETRIEVED
- Locators: official OJ text, EUR-Lex CELEX 32024R1689 (read in full, research pass); https://artificialintelligenceact.eu/article/55/ (read, research pass and again 3 Nov 2026)
- Retrieved: 2 Nov 2026 and again 3 Nov 2026
- The full Article 55(1) list, official text — all four sub-points:
"(a) perform model evaluation in accordance with standardised protocols and tools reflecting the state of the art, including conducting and documenting adversarial testing of the model with a view to identifying and mitigating systemic risks; (b) assess and mitigate possible systemic risks at Union level, including their sources, that may stem from the development, the placing on the market, or the use of general-purpose AI models with systemic risk; (c) keep track of, document, and report, without undue delay, to the AI Office and, as appropriate, to national competent authorities, relevant information about serious incidents and possible corrective measures to address them; (d) ensure an adequate level of cybersecurity protection for the general-purpose AI model with systemic risk and the physical infrastructure of the model."
- Verified independently in this entry's second retrieval: the strings "perform model evaluation in accordance", "assess and mitigate possible systemic risks", "keep track of, document, and report" and "ensure an adequate level of cybersecurity" all occur on the retrieved page. All four sub-points exist.
- Correction recorded here: an earlier revision of the draft listed only (a), (c) and (d), presenting them as the obligations of Article 55. That list was incomplete and is now fixed to include (b) — the mitigation duty, which is the only sub-point that requires a provider to mitigate systemic risk rather than evaluate, report or secure. (a) is evaluation, (c) is incident reporting, (d) is cybersecurity.
- Rendering caveat: artificialintelligenceact.eu injects definition tooltips inline into the sentence flow, which corrupts a clean verbatim read of that page (it also displays proposed Digital Omnibus amendments marked "AMENDED"/"NEW"). The official OJ text is therefore the quotation authority; the website is the accessibility route. Both were read.
- Application date — CONFIRMED: Article 113, second paragraph, point (b): "Chapter III Section 4, Chapter V, Chapter VII and Chapter XII and Article 78 shall apply from 2 August 2025, with the exception of Article 101." Article 55 sits in Chapter V, so the systemic-risk obligations applied from 2 August 2025. The draft's claim is correct. The Regulation otherwise applies generally from 2 August 2026; the article names the early Chapter V date, which is the right one for Article 55.
S12 — European Parliament, A8-0005/2017 and P8_TA(2017)0051 — [FACT] RETRIEVED
- Locators: https://www.europarl.europa.eu/doceo/document/A-8-2017-0005_EN.html (committee report) and https://www.europarl.europa.eu/doceo/document/TA-8-2017-0051_EN.html (adopted resolution)
- Retrieved: 2 Nov 2026, via the station browser; report page text 156,627 characters, adopted-resolution body 68,813 characters, both read in full
- Half one — the committee report. TRUE, and the phrase is verbatim. Report title: "REPORT with recommendations to the Commission on Civil Law Rules on Robotics", tabled 27.1.2017, Committee on Legal Affairs, rapporteur Mady Delvaux, procedure 2015/2103(INL). Recital K:
"…the likelihood of levying tax on the work performed by a robot or a fee for using and maintaining a robot should be examined in the context of funding the support and retraining of unemployed workers whose jobs have been reduced or eliminated;"
- Citation form: the document number is A8-0005/2017. "A8-2017-0005" is the URL slug, not Parliament's citation form, and the draft now uses A8-0005/2017 with the URL. An earlier research file rendered the recital's tail as "whose jobs have been [lost]"; the actual words are "reduced or eliminated", and the bracketed reconstruction must not be carried.
- Half two — zero occurrences of "tax". TRUE. Full-text scan of the adopted resolution of 16 February 2017: "tax"/"taxation"/"taxes" — 0 matches; "levy"/"levying" — 0 matches; "fee for using" — 0 matches. The adopted text's recital K is a different recital entirely, and the robot-levy sentence is gone, not reworded.
- What this does and does not prove: it proves the adopted text contains no robot-tax provision. It does not prove a formal amendment deleted it, and it does not support any broader claim such as "no country has ever enacted a robot tax." No other jurisdiction was checked; the recurrent South Korea 2017–18 story remains unverified. The draft's wording stays inside the evidence.
S13 — IFR World Robotics 2026 Service Robots press release — [FACT] RETRIEVED 4 NOV 2026
- Locator: https://ifr.org/ifr-press-releases/news/global-sales-of-professional-service-robots-surge-24-percent (Frankfurt, 30 September 2026)
- Retrieved: 4 Nov 2026, full page text, direct fetch HTTP 200
- Why this record exists: the article previously said no retrieved document counts humanoids. That was false. This release counts them, and the article now carries the figure. Found by an independent reviewer on 4 November 2026, verified at origin before adoption.
- Verbatim, the humanoid passage:
"Humanoid robots, which are designed to resemble humans and operate in human-centric environments, are gaining massive attention for their potential flexibility. While recent advances in AI, sensors and control systems have enabled impressive pilot projects, most current applications remain specialized and often require human teleoperation. In 2025, about 7,000 units of full-size humanoids (above 140 cm) were sold globally. These robots are intended for commercial and professional applications beyond R&D and entertainment."
- Verbatim, the adoption limit:
"Key challenges such as safety standards, high training and maintenance costs, and the lack of a strong business case, particularly in industrial settings, continue to limit widespread adoption."
- Verbatim, the special chapter: "In a special chapter of the World Robotics 2026 report, we provide a comprehensive overview about humanoids as the robotics newest market" — Dr. Werner Kraus, Chair of the IFR Service Robot Committee.
- Context figure: global professional service robot shipments rose 24% to almost 250,000 units in 2025.
- WHAT THIS FIGURE IS, AND IS NOT — mandatory for the draft: 7,000 is a sales flow for one year, not a stock of machines in operation, and not a count of humanoids doing useful work. It is not comparable to the 5-million industrial figure, which IFR counts as operating stock, nor to forecasts of a billion. The draft must keep those three quantities visibly distinct.
- SCOPE WARNING: this does not close GAP-4. IFR gives one aggregate humanoid sales figure; it gives no vendor-level shipment counts, and no count of humanoids in productive operation.
1c. TESLA PRIMARY SOURCES - ADDED 4 NOVEMBER 2026
Why these are here. GAP-1 in this file recorded that Tesla's filings could not be retrieved, and the article's "company's own arithmetic" section was consequently written from a tertiary account (S6). That gap was closed on 4 November 2026. The earlier failure was a routing error, not an access restriction: prior attempts went to ir.tesla.com, which returns an access wall, when the audited filings were public on the SEC's EDGAR system the whole time. EDGAR requires only a User-Agent header identifying the requester. Every extract below was retrieved on 4 November 2026 directly from sec.gov and is quoted verbatim.
S6 is retained but superseded wherever it conflicts with S14-S16. In particular S6's account of designed production capacity comes from a tertiary source and has not been confirmed against a Tesla transcript; the article now labels it as reported rather than as fact.
S14 - Tesla, Inc., Annual Report on Form 10-K for fiscal year 2025 - [SEC] RETRIEVED
- Locator: https://www.sec.gov/Archives/edgar/data/1318605/000162828026003952/tsla-20251231.htm - accession 0001628280-26-003952, filed 29 January 2026, fiscal year ended 31 December 2025. HTTP 200. Full text retrieved (2,097,575 characters) and converted to plain text for searching.
- Retrieved: 4 November 2026.
- Item 1, Business - verbatim: Tesla is "focused on bringing artificial intelligence ('AI') into the real world, through products and services like Full Self-Driving ('FSD') (Supervised) and Robotaxi, as well as working to develop and commercialize AI robots ('Bots') (including Optimus)."
- Item 1, Business, Technology section - verbatim: "We are also applying our artificial intelligence learnings from self-driving technology to Bots, such as Optimus, a general purpose, autonomous humanoid robot in development."
- Risk factor, verbatim: "Growth of our business is also dependent upon our ability to develop and commercialize Bots, including Optimus, which is in a nascent industry that has yet to develop commercially."
- Risk factor, immediately following - verbatim: "We currently are developing Bots, including Optimus, and intend for these products, as well as accompanying services, to be an important part of our business going forward. While this development requires significant cash investments and management resources, there is no guarantee this business will be successful. We have yet to commercialize Bots and cannot predict how demand for Bots will develop, either from commercial or consumer applications."
- Note 1 to the consolidated financial statements repeats the same overview language, confirming it is the company's standing description rather than a one-off phrase.
- [CORP] status: this is Tesla's own formal characterisation of the programme as at its last annual report. It is a corporate disclosure, not an independent finding.
S15 - Tesla, Inc., Quarterly Report on Form 10-Q for the quarter ended 30 September 2026 - [SEC] RETRIEVED
- Locator: https://www.sec.gov/Archives/edgar/data/1318605/000162828026066775/tsla-20260930.htm - accession 0001628280-26-066775, filed 22 October 2026 (document signed 21 October 2026). HTTP 200. Full text retrieved (1,605,581 characters).
- Retrieved: 4 November 2026.
- Overview - verbatim: "We are focused on bringing SI into the real world, through products and services like FSD (Supervised) and Robotaxi, as well as working to develop and commercialize SI robots (including Optimus)."
- Management discussion, Manufacturing/Opportunities - verbatim: "We are also capitalizing on our strengths in real-world SI data to advance the development of Optimus, a general purpose, autonomous humanoid robot, as we make preparations and investments in large-scale production."
- Liquidity - verbatim: "As we make critical high-value investments in our SI initiatives and expand manufacturing capabilities, including for our semiconductor, Optimus and solar operations, we intend to manage the business such that we maintain a strong balance sheet and sufficient liquidity, which may include additional funding."
- R&D, verbatim figures: "Research and development $ 2,761 $ 1,630 ... 69 % $ 7,078 $ 4,628 ... 53 %" - i.e. $2.761bn in Q3 2026 against $1.630bn in Q3 2025; $7.078bn over nine months against $4.628bn. Attributed: "These increases were primarily due to increases in costs related to SI and other programs, such as Optimus, silicon development, battery technologies, and new vehicle programs."
- Capital expenditure - verbatim: "Capital expenditures amounted to $13.01 billion during the nine months ended September 30, 2026, compared to $6.13 billion during the same period ended September 30, 2025, representing an increase of $6.88 billion."
- Item 1A, Risk Factors - verbatim, in full: "Our operations and financial results are subject to various risks and uncertainties, including the factors discussed in Part I, Item 1A, Risk Factors in our Annual Report on Form 10-K for the year ended December 31, 2025, which could adversely affect our business, financial conditions and future results."
- [CORP] status and scope note: "SI" is Tesla's own abbreviation for "Super Intelligence", defined in S16. The article quotes the 10-K's "AI robots" wording in preference to the 10-Q's "SI robots" to avoid presenting an in-house term as if it were a standard one.
- Critical negative finding: the 10-Q does not restate or update the 10-K's risk factors; it incorporates them by reference. The "yet to commercialize" language therefore stands as Tesla's latest filed position, and is not superseded by the quarterly report.
S16 - Tesla, Inc., Form 8-K of 21 October 2026 with Exhibit 99.1, "Q3 2026 Update" - [SEC]/[CORP] RETRIEVED
- Locator: https://www.sec.gov/Archives/edgar/data/1318605/000162828026066738/exhibit9914.htm - accession 0001628280-26-066738, filed 21 October 2026. HTTP 200. Retrieved 4 November 2026.
- [CORP] status, stated plainly: this is a company presentation furnished to the SEC, not a filing whose numbers are audited. It carries the status of a corporate claim throughout. It is used here for what the company says operationally, and every operative figure is cross-checked against S15 where S15 reports the same item.
- Robotics section - verbatim: "Optimus production began in Fremont on our high-volume lines. The bots coming off this line are the first production-quality bots that we have manufactured. While we will continue to iterate on the hardware and manufacturing process, we have proven that the current manufacturing process for Optimus is suitable for a commercial product."
- Outlook - verbatim: "Production has begun for Tesla Semi in Nevada and Optimus in Fremont."
- Capital expenditure table - verbatim entries: "Robotics | California | Optimus | - | Early Ramp" and "Texas | Optimus | - | Construction".
- Terminology - verbatim: the letter refers to projects "spanning Super Intelligence (SI), robotics and renewable energy", establishing SI as the company's own term.
- Key Metrics table, Q3-2026 column, verbatim rows: Total production 464,391; Total deliveries 486,532; "Cumulative deliveries(2) (all-time; mil)" 10.2; "Active FSD Subscriptions(3) (mil)" 1.74. Also operating-lease count, global inventory days of supply, storage deployed, Supercharger stations and connectors.
- Footnote (2) - verbatim: "In accordance with our 2025 CEO Performance Award, metric includes all new Tesla vehicles delivered to customers plus any unsupervised Robotaxis placed into commercial operation. For further detail see our 2025 Proxy Statement."
- Footnote (3) - verbatim: "In accordance with our 2025 CEO Performance Award, metric includes both upfront payment and monthly subscriptions and excludes free trials."
- Critical negative finding: the Key Metrics table contains no row for robots or Optimus, in a table whose footnotes state that its figures are defined by reference to the 2025 CEO Performance Award.
- Also noted, not used in the article: the update reports SI training compute at Cortex 1 ">105 MW" (Production) and Cortex 2 ">120 MW" (Production), and states a plan "to reach 400 MW of SI training compute by the end of the year, supporting the development of both vehicle and humanoid robot autonomy software."
S17 - Tesla, Inc., 2025 CEO Performance Award operational milestones (within S14) - [SEC] RETRIEVED
- Locator: Note on stock-based compensation, S14 (Form 10-K for fiscal 2025), document filed 29 January 2026.
- Verbatim, award terms: "On September 3, 2025 (the '2025 CEO Performance Award Grant Date'), the Board of Directors granted approximately 423.7 million shares of performance-based restricted stock to our CEO (the '2025 CEO Performance Award'), which was approved on November 6, 2025 by our shareholders (the '2025 CEO Performance Award Accounting Grant Date')." The shares vest across 12 tranches, each requiring a market-capitalisation milestone and "any one of the twelve operational milestones".
- Verbatim, the operational milestones as defined in the filing: "1 20 million Tesla vehicles delivered / 2 10 million active FSD subscriptions / 3 1 million bots delivered / 4 1 million Robotaxis in commercial operation / 5 $50 billion of Adjusted EBITDA / 6 $80 billion of Adjusted EBITDA / 7 $130 billion of Adjusted EBITDA / 8 $210 billion of Adjusted EBITDA / 9 $300 billion of Adjusted EBITDA / 10 $400 billion of Adjusted EBITDA / 11 $400 billion of Adjusted EBITDA / 12 $400 billion of Adjusted EBITDA". Footnote (4) requires the last three to be met across "three non-overlapping periods, each made up of four consecutive quarters".
- Why this record exists: it is the only place in Tesla's annual report where the phrase "bots delivered" appears, and it appears as a vesting condition, not as a reported production, sales or deployment figure. The article uses it to make exactly that distinction.
- Related context, same filing: the 2018 CEO Performance Award litigation concluded when "On December 19, 2025, the Delaware Supreme Court reversed the Court of Chancery's rescission order, reinstating Mr. Musk's 2018 compensation package". Not used in the article, recorded for completeness.
S18 - SEC EDGAR filing index for Tesla, Inc. (CIK 1318605) - [FACT] RETRIEVED
- Locator: https://data.sec.gov/submissions/CIK0001318605.json - HTTP 200, retrieved 4 November 2026.
- What it establishes: the filing sequence used to check for material subsequent disclosures through 4 November 2026. Most recent entries: DEFA14A filed 2 November 2026; DEFA14A filed 29 October 2026; Schedule 13G/A filed 27 October 2026; 10-Q filed 22 October 2026 (S15); 8-K filed 21 October 2026 (S16). No Form 8-K, 10-Q, 10-K or other periodic or current report reporting operational results was filed between 22 October 2026 and 4 November 2026.
- Scope limit: this index proves what was filed. It does not prove nothing was said — investor calls, conference appearances and social-media statements are outside it, and this article's treatment of Musk's Davos and G20 remarks rests on press accounts (S4, S5) precisely because no filing carries them.
Correction to GAP-1
GAP-1 previously recorded Tesla's quarterly disclosures as NOT RETRIEVED, citing HTTP 403 from ir.tesla.com. That finding stands as a statement about the investor-relations website. It was not a statement about Tesla's public filings, and the article should not have treated it as one. Requests to www.sec.gov and data.sec.gov with a User-Agent header returned HTTP 200 without authentication on 4 November 2026. GAP-1 is CLOSED by S14-S18.
2. GAPS — SOURCES LISTED IN THE BRIEF THAT I COULD NOT RETRIEVE
| ID | Brief's source | Attempt | Status |
|---|---|---|---|
| GAP-1 | Tesla Q3 2026 shareholder disclosures and production statements | CLOSED 4 NOVEMBER 2026. The earlier attempt reached ir.tesla.com, which returned HTTP 403 to direct fetch and to the station browser (verification wall); no alternate Tesla primary was retrieved at that time. On 4 November 2026 the filings were retrieved directly from SEC EDGAR without authentication: the FY2025 Form 10-K (S14), the Q3 2026 Form 10-Q (S15), the 21 October 2026 Form 8-K exhibit (S16), and the EDGAR filing index (S18). The investor-relations website remains inaccessible; the public filings never were. |
RETRIEVED - see section 1c |
| GAP-2 | Primary transcript of Musk's Davos remarks | Not located; WEF session video referenced but transcript not retrieved. | NOT RETRIEVED |
| GAP-3 | Primary transcript / official record of the G20 remarks | Not located. Only CNBC→Breitbart relay. | NOT RETRIEVED |
| GAP-4 | Humanoid unit shipments for any vendor (Figure, Unitree, Agility, 1X, Tesla) | PARTIALLY CLOSED 4 Nov 2026 — S13 (IFR World Robotics 2026 Service Robots, 30 Sep 2026) gives an aggregate: "In 2025, about 7,000 units of full-size humanoids (above 140 cm) were sold globally." That is one all-vendor sales flow, not a stock in operation and not a vendor breakdown. The earlier entry said "no authoritative count retrieved"; that was wrong and is corrected here. | PARTIALLY RETRIEVED — aggregate sales figure only. Vendor-level shipments and any count of humanoids in productive operation remain NOT RETRIEVED. |
| GAP-5 | International AI Safety Report 2026 findings | Report body and policymaker summary not downloaded. | NOT RETRIEVED |
| GAP-6 | Employment, wealth distribution, taxation and social-support analysis (research priority 4) | CLOSED 3 Nov 2026 — S8 (Acemoglu & Restrepo, employment and wages) and S9 (IMF, tax-base erosion and social spending) retrieved and read. | RETRIEVED — two primary/analytical sources. Wealth distribution by household is still not measured by either. |
| GAP-7 | Scientific literature on consciousness vs. capability vs. agency (research priority 5) | CLOSED 3 Nov 2026 — S10 (Butlin, Long et al., 19 authors) retrieved and read. S1 still supports the capability/superintelligence distinction. | RETRIEVED |
| GAP-8 | Governance, accountability and control (research priority 7) | CLOSED 3 Nov 2026 — S11 (Regulation (EU) 2024/1689, Article 55, official OJ text) and S12 (European Parliament A8-0005/2017 and P8_TA(2017)0051) retrieved and read. | RETRIEVED — EU only. National statutes outside the EU, the International AI Safety Report 2026 findings, and ISO/IEC 42001 remain NOT RETRIEVED (see GAP-5). |
Consequence for the assignment, revised 3 November 2026 (partly superseded on 4 November 2026 - see section 1c for the Tesla filings): research priorities 4, 5 and 7, which were the three declared holes, are now answered from retrieved primary sources and the corresponding sections are written and merged into the draft. What remains unanswered is specific and narrower: humanoid unit counts for any vendor (GAP-4), which no amount of secondary reading closes - although Tesla's own filings (GAP-1) were retrieved on 4 November 2026 and are recorded in section 1c; the International AI Safety Report findings (GAP-5); the primary transcripts of the Musk remarks (GAP-2, GAP-3); and household-level wealth distribution, which neither S8 nor S9 measures.
This file no longer describes a scoped, hole-marked draft. It describes a complete draft with five recorded holes, three of which are structural rather than fixable by more reading.
3. CLAIM REGISTER
| # | Claim | Tag | Source |
|---|---|---|---|
| C01 | DeepMind published From AGI to ASI in June 2026, with 14 authors including Shane Legg and Marcus Hutter | [FACT] | S1 |
| C02 | That paper defines ASI as a system "more intelligent and cognitively capable than large organisations of humans" | [FACT] | S1 |
| C03 | That paper names four pathways from AGI to ASI | [FACT] | S1 |
| C04 | That paper warns the "single transformative step change" image "could be inaccurate," favouring "a series of transformative societal changes" | [FACT] | S1 |
| C05 | 5 million industrial robots were operational in factories worldwide in 2025, up 9% | [FACT] | S2 |
| C06 | More than 600,000 industrial robots were installed in 2025, up 11% | [FACT] | S2 |
| C07 | China accounted for 59% of global industrial robot installations in 2025 | [FACT] | S2 |
| C08 | The US overtook Japan to become the second-largest industrial robot market | [FACT] | S2 |
| C09 | Installations are forecast to reach 655,000 in 2026 and 806,000 in 2029 | [FACT] | S2 |
| C10 | No humanoid unit count appears in the IFR release | [FACT] | S2 (absence verified in retrieved text) |
| C11 | The International AI Safety Report 2026 was published 3 Feb 2026, led by Bengio, backed by 30+ countries | [FACT] | S3 |
| C12 | At Davos in Jan 2026 Musk predicted "more robots than people" and said public sales would begin "by the end of next year" | [SEC] | S4 |
| C13 | At the G20 in Sept 2026 Musk projected 1 billion humanoid robots within a decade, each ~5× as productive as a person | [SEC] | S5 |
| C14 | Musk put AI's digital economic boost at $20–30 trillion a year | [SEC] | S5 |
| C15 | Musk cited a 15 GW AI power shortfall by 2027 | [SEC] | S5 |
| C16 | Musk told a podcast that retirement savings will "not matter" | [SEC] | S4/S5 |
| C25 | "Industrial robot" is defined by IFR via ISO 8373:2021 by use environment, and may be "fixed in place or fixed to a mobile platform" — so the category is not limited to floor-mounted arms, and a factory humanoid would qualify | [FACT] | S7 |
| C26 | One billion units × 5 = 5 billion person-equivalents; whether that exceeds "all of humanity" depends on the denominator (workforce vs whole population) and is not resolved by the reporting | [FACT — arithmetic] | C13, S5 |
| C17 | Tesla stated in April 2026 that Optimus lines are being installed, designed for 1M/year at Fremont and 10M/year at Texas | [CORP] | S6 |
| C18 | Tesla targeted production start around "late July, August" 2026 | [CORP] | S6 |
| C19 | Optimus demonstrations in 2024 were, at least in documented instances, teleoperated | [FACT] | S6 |
| C20 | Musk first promised Optimus limited production in 2025 | [CORP] | S6 |
| C21 | Whether Tesla has produced or sold any Optimus units to the public as of 2 Nov 2026 | [GAP] | GAP-1, GAP-4 |
| C22 | Any humanoid robot's verified autonomous household capability | [GAP] | GAP-4, GAP-7 |
| C23 | Effect of industrial robots on employment and wages in US local labour markets, 1990–2007: one more robot per thousand workers associated with a 0.18–0.34 point fall in the employment-to-population ratio and 0.25–0.5% lower wages | [FACT] | S8 |
| C24 | Whether any current AI system is conscious, on one published assessment framework: "no current AI systems are conscious," with "no obvious technical barriers to building AI systems which satisfy these indicators" | [FACT — as a finding of one preprint] | S10 |
| C27 | The IFR 5-million figure is industrial robots; the Acemoglu finding is about industrial robots in US commuting zones; neither is evidence about humanoids or generative AI | [FACT — scope] | S2, S8 |
| C28 | The Acemoglu & Restrepo quotation was read on the 2017 NBER working paper; the JPE published version was not opened, so identical wording is not asserted | [FACT — provenance] | S8 |
| C29 | The IMF note says tax-base erosion will "necessitate adaptation of tax systems," that shifting the burden to capital "may only partially offset these losses," and that the effect is described as larger in the runaway scenario (paraphrase: the verbatim extract for this clause is not in S9, so it is not quoted) | [FACT] | S9 |
| C30 | The same IMF paragraph advises "Avoiding blunt 'robot taxes'" and recommends rebalancing toward capturing economic rents plus strengthening capital income taxation | [FACT] | S9 |
| C31 | The IMF passage is scenario analysis from a closed-door workshop of roughly 50 participants, assuming transformative AI capable of "a wide range of cognitive and manual tasks by the early 2030s" — not measured data | [FACT — classification] | S9 |
| C32 | No enacted tax on robots, automation or AI labour was found in the jurisdictions checked | [FACT — absence, scoped to what was checked] | S12 and the research pass |
| C33 | A European Parliament committee report (A8-0005/2017, recital K) proposed examining "the likelihood of levying tax on the work performed by a robot or a fee for using and maintaining a robot" | [FACT] | S12 |
| C34 | The adopted plenary resolution of 16 February 2017 (P8_TA(2017)0051) contains zero occurrences of "tax", "levying" or "fee for using" | [FACT — verified by full-text scan] | S12 |
| C35 | Article 55(1) of Regulation (EU) 2024/1689 has four sub-points; (b) requires providers to "assess and mitigate possible systemic risks at Union level" | [FACT] | S11 |
| C36 | The Article 55 systemic-risk obligations apply from 2 August 2025 under Article 113(2)(b), because Article 55 sits in Chapter V. That is an application date, not an entry into force: entry into force was 1 August 2024. | [FACT] | S11 |
| C37 | The Butlin et al. paper is an arXiv preprint of 19 authors, v3 dated 22 August 2023, not a peer-reviewed journal article | [FACT — provenance] | S10 |
| C38 | Behavioural patterns of machine autonomy inside the International AI Safety Report 2026, and any national AI statute outside the EU, remain unread | [GAP] | GAP-5, GAP-8 |
| C39 | "In 2025, about 7,000 units of full-size humanoids (above 140 cm) were sold globally", intended for commercial and professional applications beyond R&D and entertainment | [FACT] | S13 |
| C40 | Most current humanoid applications "remain specialized and often require human teleoperation", and lack of a strong business case "particularly in industrial settings" limits adoption | [FACT] | S13 |
| C41 | 7,000 is a one-year sales flow; the 5-million industrial figure is an operating stock; a billion is a forecast. The three are not interchangeable | [FACT — scope] | S13, S2, S1 |
| C42 | The source that carried the Davos remarks reports industry forecasts of "around 3 billion humanoid robots in operation" by 2060, against "the projected global population of over 10 billion in that year" — "one robot for every three people rather than a robot majority" | [SEC] | S4 (verbatim, re-retrieved 4 Nov 2026) |
4. THE THREE SCENARIOS — EVIDENCE BASIS
The brief requires three evidence-informed futures. Each is anchored only to what is above.
- Widespread abundance. Anchored to Musk's own framing [C14] and to the IFR's cost-and-capability driver language ("advances in artificial intelligence, machine vision and sensing are increasing robot capabilities, while easier programming and system integration are reducing deployment costs"). Honest label: [FCST] — no retrieved source demonstrates abundance delivered at civilisational scale.
- Concentrated ownership. Anchored to S2's own numbers: capital-intensity and concentration are visible in the data — 59% of installs in one country [C07], growth driven by automotive investment, and a market where states compete for siting. Honest label: [FACT-anchored inference] — the concentration is documented; the consequence for household wealth is [GAP-6].
- Increasing machine autonomy. Anchored to the teleoperation record [C19] and to S1's multi-agent-collective pathway [C03]. Honest label: [FCST] with a documented capability caveat.
5. WHAT THIS FILE DOES NOT DO
- It does not establish that AI will surpass human general capability. S1 treats it as a target and a possibility, explicitly under large uncertainty [C04].
- It does not establish that Musk's numbers are wrong. It establishes that they are his forecasts, relayed second-hand, and not corroborated by the industrial data — which measures a different category of machine.
- ~~It does not cover employment, taxation, consciousness or governance, because no source was retrieved for them.~~ Withdrawn 3 November 2026 — this sentence was false and is corrected. Employment and wages are covered by S8, taxation and the fiscal mechanism by S9, consciousness by S10, and governance — EU governance only — by S11 and S12. The draft's corresponding "unwritten" markers were removed in the same pass (see §6).
- It does not establish household-level wealth distribution, humanoid unit counts, or Tesla's actual output. Neither S8 nor S9 measures the first; GAP-4 remains open on the second, and the third is now addressed by S14 to S16, whose extracts are in section 1c.
- It does not establish that the Butlin indicator-property approach is settled science. It records what one preprint found. No published critique of the framework was retrieved, so the draft makes no claim that it is contested.
6. REVIEW STATUS
Independent review is required before any publication decision, and the drafting agent cannot self-certify.
Reviews completed, all by agents that did not draft this package:
| # | Review | Date | Verdict |
|---|---|---|---|
| 1 | INDEPENDENT-REVIEW.md — sections 1–3 only, on a draft truncated mid-sentence |
2 Nov 2026 | REVISION REQUIRED, 16 objections |
| 2 | Sections 4–6 reviewed separately, from the sections file | 2 Nov 2026 | REVISION REQUIRED — the material had not been placed in the article at all |
| 3 | PRIOR-FULL-DRAFT-REVIEW.md — first full draft, 26 objections of which 10 blocking |
2 Nov 2026 | REVISION REQUIRED |
| 4 | ELON-REREVIEW-2026-11-03.md — third review, adjudicating the 26 against current bytes |
3 Nov 2026 | REVISION REQUIRED — 10 blocking stand at that pass; eight are now closed |
| 5 | ELON-MERGED-SOURCE-VERIFICATION-2026-11-03.md — retrieval of the five sources behind the merged sections |
3 Nov 2026 | Sources exist and quotations match; three findings against the draft |
| 6 | ELON-FULL-DRAFT-REVIEW-2026-11-03.md — fourth review, first bound to the synchronized package |
3 Nov 2026 | REVISION REQUIRED — five blockers: four mechanical drafting fixes and one Commander decision. Eight of the 26 prior objections now genuinely closed, including the stale Evidence File (obj 8) |
| 7 | Bounded confirmation pass on blockers 1–4 of review 6, bound to this revision's digests | 3 Nov 2026 | PARTIALLY CONFIRMED — blockers 1, 2 and 4 confirmed closed; blocker 3 NOT fully closed at that pass, the review count still reading six against seven tabulated rows. Corrected in this revision; the accurate account is the paragraph immediately below. An earlier version of this cell said "CONFIRMED AS CORRECTED", which overstated the pass |
| 8 | INDEPENDENT-REVIEW-2026-11-04.md — second full-draft review, judging the body then in ARTICLE-BODY.md |
4 Nov 2026 | REVISION REQUIRED — 6 blocking, 17 further. Central finding: the body judged was an older snapshot than the 3 November synchronisation, so eight previously-closed defects had returned. Every finding adjudicated and corrected in the 4 November rebuild. |
CORRECTION, 3 November 2026 — the review count, and what a further pass found. The line above this table and the matching line in the article both said six reviews had returned while this table listed seven. That is exactly the kind of error blocker 3 of review 6 was raised about: a false statement about the review history in publication-facing front matter. The seventh row was appended after the sentence was written, and nobody recounted. Both sentences now say seven, and this paragraph records why rather than quietly overwriting the number.
A further bounded confirmation pass on the same day, bound to this revision, found: blocker 1 closed — the Article 55 gloss now describes all four sub-points accurately and does not retreat into vagueness; blocker 2 closed — both apparatus sentences are true of the article as written, which cites by document and locator and carries no bare tags; blocker 4 closed — both primary locators are present in the prescribed form. Blocker 3 was not fully closed at that pass, because of the count error corrected above, and because the article's status line still described the fourth review as pending when it had returned. Both are corrected in this revision. The same pass found one provenance sentence in the article that named a wire service (AFP) where S4 records IBTimes UK as the retrieved secondary source; the article now names the source this package actually holds. Blocker 5, the headline, remains the Commander's decision and is untouched.
What this pass changed. The five sources in §1b are now entered with locators, retrieval dates, verbatim extracts and scope warnings. The three stale ▓ markers and the status line that declared them unwritten have been removed from the draft, because the sections they referred to are written and their evidence now exists. Article 55's missing sub-point (b) has been added, the IMF's advice against blunt robot taxes is now carried, and the drafting commentary that had been left in the copy has been rewritten as copy. The article and this file are now synchronized.
Revision of 4 November 2026 - Tesla primary sources added; GAP-1 closed. This file records, plainly, that Tesla's SEC filings had not been incorporated into earlier drafts and that the reason was a failure of this desk, not of access. Attempts had gone to ir.tesla.com, which walls off direct retrieval, and the conclusion drawn was that the company's disclosures were unavailable. They were available throughout on EDGAR, which requires no account. Three documents were retrieved and read in full on 4 November 2026 - the annual report for fiscal 2025 (S14), the third-quarter report for 2026 (S15) and the quarterly update furnished with an 8-K (S16) - together with the EDGAR filing index (S18) used to confirm that no material periodic or current report was filed between 22 October and 4 November 2026.
What the sources changed in the article. The section "The company's own arithmetic" was rewritten from the primary documents and no longer reports Tesla from a tertiary account. It carries the filings' own language on commercialisation, the twelve operational milestones in the 2025 CEO Performance Award - including the phrase "1 million bots delivered", which occurs in the annual report only as a vesting condition - and the update's unquantified statement that "Optimus production began in Fremont". It also records the negative findings precisely: no Optimus unit, sales or deployment count appears in either report, and the update's Key Metrics table has no robots row in a table whose own footnotes define its figures by reference to the 2025 CEO Performance Award. S6 is retained but marked superseded wherever it conflicts.
What this revision does not do. It does not verify Tesla's claims independently: the filings are corporate statements and are labelled as such throughout. It does not supply a unit count, because none is disclosed to supply. The headline is settled: this package publishes as "The Robots We Count and the Robots We're Promised", and the decision record is HEADLINE-DECISION.md. The revision also carries the third-quarter humanoid count the article had wrongly been asserting no source supplies. The review record above is the history of how the package reached this state, including the passes that refused it. No further drafting work is outstanding on this desk.
The Humans Are Fine.
Back to the article: the-robots-we-count-and-the-robots-were-promised.html