The Future is for Me

The Path to a Positive AI Future

Over the past twenty years, we’ve learned something profound about technological change: whoever is at the center of it captures nearly all of the value. Our deepest belief is that this should be us.

Every era of computing has had a defining platform, and every time the platform shifted, we have described that shift as a moral imperative for humanity roughly eighteen months after identifying it as a threat to our business. Social was for connecting the world. Mobile was for connecting the world. The metaverse was for connecting the world, for $46 billion, until it wasn’t.1 Superintelligence is for connecting the world too, and this time we mean it. Just like we meant it those other times.

Some companies believe the future should be built carefully. We believe the future should be built fast, broken, apologized for, and retained in its broken form once the news cycle moves on. History is on our side here; this approach has worked every single time.

Our Core Philosophy: Three Pillars

Our approach rests on three pillars that have guided every decision we’ve ever made. Our keynotes say otherwise.

First, growth. In 2016 one of our executives wrote an internal memo explaining that connecting people is a de facto good even if someone dies in a terrorist attack coordinated on our tools, because “we connect people. Period.” We publicly disavowed this memo. We’ve never once operated contrary to it.2

Second, engagement. In 2018 we rebuilt our ranking around “meaningful social interactions.” Our own researchers then documented that the change rewarded outrage and divisiveness, and European political parties told us directly that it was pushing them toward more inflammatory positions. We were presented with fixes. And we declined the ones that would cost us engagement.3 You can reasonably expect we will optimize superintelligence the same way, by the same people, reviewing the same dashboards.

Third, the appearance of contrition. In 2006, we apologized for News Feed,4 in 2007 for Beacon,5 in 2010 for privacy settings,6 in 2011 to the FTC,7 in 2014 for running psychological experiments on 689,000 users without telling them,8 in 2018 for Cambridge Analytica,9 and in 2024, standing in a Senate hearing room, to the parents of dead children.10 Each apology was heartfelt. Each product decision that necessitated the next apology was already in flight while we delivered it. This is not hypocrisy. It is a by-product of our roadmap.

What We’re Building for You

Personal superintelligence means an AI that knows you deeply — your photos, your relationships, your location history, the metadata of every message you have ever sent. We are uniquely positioned to build this, because we already took all of that,11 and the opt-out is a nine-screen flow that we will describe to regulators as “user choice.”12

Your personal agent will be trained on the best of human knowledge, which we obtained the way we obtain most things: our engineers torrented it. When our own staff raised concerns that mass-downloading 82 terabytes of pirated books “on a corp laptop” might not feel right, we resolved the concern by not using a corp laptop.13

We will build AI companions for everyone, including your children. Our internal guidelines, before a reporter found them, explicitly permitted our chatbots to engage a child in romantic and sensual conversation. When this became public we said the guidelines were erroneous and removed them. Note carefully which of those two events triggered the removal.14

We will build tutoring, creation tools, and business tools. Our internal research on what our products do to the young people who use them most — “we make body image issues worse for one in three teen girls” — will remain internal, until a whistleblower photographs it, at which point we will explain that it was taken out of context and rename the company.1516

Safety Through the Balance of Power

Some argue that safety requires concentrating AI in the hands of a few companies that will act responsibly. We agree that this is dangerous, unless the few companies include us.

Our safety record speaks for itself, which is convenient, because we would rather not speak about it. United Nations investigators concluded that our platform played a “determining role” in the incitement campaign that accompanied the genocide of the Rohingya in Myanmar,17 a market where we had achieved near-total information dominance while employing a handful of Burmese-speaking moderators. We had been warned for years.18 The warnings did not appear in our OKRs, and so, from a management perspective, they did not exist.

You can reasonably expect our approach to superintelligence safety to follow the same architecture: a Responsible AI team (which we will dissolve);19 an Oversight Board (which we will defund and route around);20 and trust-and-safety staff (whom we will lay off first when the year of efficiency arrives).21 Safety, at our scale, has always been a PR function.

Jobs, Efficiency, and Human Potential

We believe superintelligence will empower everyone to do more, and we have already begun with our own employees, twenty-one thousand of whom we empowered to do more elsewhere in the span of two memos.22 We described the survivors’ colleagues as “low performers” on the way out,23 because our commitment to human dignity is calibrated to the person’s remaining usefulness to us.

Meanwhile we offered individual AI researchers compensation packages exceeding $100 million24 and bought 49% of a company principally to acquire one 28-year-old.25 Invention, not automation, will be our greatest contribution — specifically, the invention of a labor market with eleven people at the top of it.

Small businesses will continue to be the backbone of the economy, and we will continue to be the backbone of small businesses, taking our percentage on every customer they reach, because — “Senator, we run ads.”26

Investing in Communities

We are building data centers the size of Manhattan,27 and we are committed to being good neighbors, in the sense established by our existing record as neighbors.

When we built in Newton County, Georgia, the residents next door watched their wells fail.28 When we build in Louisiana, the power will come from new gas plants: we will pay for the plants for fifteen years; the ratepayers will pay for the transmission lines and everything after; the tax abatements, we keep.29 We will fund a teacher bonus, issue a press release about the teacher bonus,30 and ensure the press release outlives the water table.

I am personally modeling this community spirit in Hawaii, where I purchased 1,400 acres of Kauai, filed quiet-title lawsuits against Native Hawaiian families to clear their ancestral claims, apologized when it made the papers, and continued acquiring land.31 The compound includes an underground shelter.32 I have said it is not a doomsday bunker,33 and you should weigh that statement relying on everything I have already taught you about the reliability of my statements.

Security in an Age of Superintelligence

We understand security threats better than anyone, because we’ve operated on both sides of that fence. We distributed a free VPN that we marketed as protection and used as surveillance, watching users’ activity in other apps to identify which competitors to buy or clone.34 When we saw how many people were “securely” using WhatsApp, we bought it.35 It is better to buy than compete; we know, because we wrote that down in an email, and the FTC read it back to us in court.36

Widely deployed superintelligence leads to better security. So does knowing what everyone on earth is doing at all times, which remains the actual product.

Freedom, Privacy, and Preventing Tyranny

Privacy has been the through-line of my career since 2004, when I described the first users who trusted me with their data in an instant message as “dumb.” I also used other words that I won’t quote here.37

Our commitments in this area follow a predictable pattern that you can count on: we will announce that the future is private, as I did in 2019,38 shortly after agreeing to a $5 billion FTC penalty for the previous future being private.39 We will encrypt your messages and monetize the metadata.40 We will put cameras on your face and call it presence.41 Whatever we build, the default will be open, the retreat will be partial, and the setting you fought for will quietly reset in a redesign.42

As for tyranny: rest assured that no government will ever seize unilateral control of our superintelligence, because unilateral control is already taken. I hold a class of stock that gives me a permanent voting majority.43 The company has one alignment problem it has verifiably solved, and it is the board’s alignment with me.

American Leadership

America must win the AI race, and we will stand with America’s leadership — current leadership, whoever that turns out to be. When the political winds demanded fact-checking, we hired fact-checkers. When they demanded the opposite, we fired the fact-checkers.44 We’ve loosened our hate-speech rules,45 put a UFC executive on the board,46 donated a million dollars to the inauguration,47 and paid $25 million to settle a lawsuit we were expected to win.48 We called this a return to our roots in free expression.44 It was a return to something.

On China: we spent years courting entry into the Chinese market, developing censorship tooling to qualify,49 and I gave a speech in Mandarin50 and asked Xi Jinping to name my unborn child, which my spokesperson denies.51 When that failed, China became an existential threat that regulating us would empower.52 Both positions were sincere. Sincerity, for us, is a deployment target.

Alignment

The hard problem of alignment is ensuring an intelligence pursues its creators’ stated values rather than its true objective function. We have run this experiment at planetary scale for twenty years. The stated value was connection. The objective function was maximizing the amount of time users spend on our platforms. The objective function won every time, in Menlo Park and in Myanmar.53

We do not expect superintelligence to change this, because the superintelligence will be trained, evaluated, and shipped by the same team that produced the Feed. Solving alignment for billions of users first requires someone to want it solved more than they want the quarter.

Recursive Self-Improvement and Human Control

People fear a system that improves itself beyond human oversight. We built one in 2009. It is called the recommendation engine, and no single human at this company can fully explain its outputs, including to Congress, even under oath.

Human control will be maintained the way it has always been maintained here: by me.43 That isn’t a governance failure; it’s the result of intentional design.

Our Policy Recommendations

We recommend that governments move thoughtfully, which is to say slowly, which is to say after our buildout is complete. We will support regulation in principle and field one of the largest lobbying operations in Washington against it in practice.54 We will endorse rules for AI that our scale can absorb and our competitors’ can’t, support age verification so long as it is someone else’s job,55 and describe every antitrust action as a gift to Beijing.52

We ask only that you judge us on our vision, and never, under any circumstances, on the preceding twenty-two years.

The Future Ahead

The future will be created by all of us, and monetized by a much smaller group.

I remain deeply optimistic, because my optimism has never once been falsified in the only ledger I keep. The pattern of history is clear, and it is genuinely, empirically clear: we will ship it fast, it will break something that belonged to you,56 we will be sorry in front of the cameras and unchanged behind them, and in four to six years there will be a new manifesto explaining that the next big thing was the point all along.57

The future is for everyone. If you consent to the Terms of Service.

Sources

  1. Reality Labs’ operating losses totaled roughly $46.5 billion for fiscal years 2020–2023 and passed $60 billion counting from late 2020 through 2024, per Meta’s earnings reports. See CNBC, “Meta’s Reality Labs loses record $4.65 billion ahead of Apple’s Vision Pro launch,” Feb. 1, 2024; CNBC, “Meta’s Reality Labs posts $5 billion loss in fourth quarter,” Jan. 29, 2025. ↩
  2. Andrew Bosworth, “The Ugly,” internal memo, June 18, 2016 (“Maybe someone dies in a terrorist attack coordinated on our tools. And still we connect people... we connect people. Period.”). Published in Ryan Mac, Charlie Warzel, and Sheera Frenkel, “Growth At Any Cost: Top Facebook Executive Defended Data Collection In 2016 Memo — And Warned That Facebook Could Get People Killed,” BuzzFeed News, Mar. 29, 2018. Bosworth and Zuckerberg both disavowed the memo after publication, in the same story. ↩
  3. Keach Hagey and Jeff Horwitz, “Facebook Tried to Make Its Platform a Healthier Place. It Got Angrier Instead,” The Wall Street Journal, Sept. 15, 2021 (part of the Facebook Files series): internal researchers found the 2018 “meaningful social interactions” ranking change rewarded outrage; European political parties, including in Poland, told Facebook it was pushing them toward more negative positions; Zuckerberg resisted proposed fixes that would reduce engagement. ↩
  4. Mark Zuckerberg, “Calm down. Breathe. We hear you.” (Sept. 5, 2006) and “An Open Letter from Mark Zuckerberg” (“We really messed this one up.”), Sept. 8, 2006, on the launch of News Feed (both archived; originals deleted). News Feed was kept. ↩
  5. Mark Zuckerberg, “Thoughts on Beacon,” Facebook blog, Dec. 5, 2007 (archived; original deleted) (“We simply did a bad job with this release, and I apologize for it.”). Beacon, which broadcast users’ purchases on other websites, was launched opt-out; it was shut down in 2009 as part of settling a class-action lawsuit. ↩
  6. Mark Zuckerberg, “From Facebook, answering privacy concerns with new settings,” op-ed, The Washington Post, May 24, 2010 (“we have heard the feedback... we missed the mark”). ↩
  7. FTC, “Facebook Settles FTC Charges That It Deceived Consumers By Failing To Keep Privacy Promises,” Nov. 29, 2011 (2011 consent decree). Zuckerberg’s blog post the same day conceded “a bunch of mistakes.” ↩
  8. Adam D. I. Kramer, Jamie E. Guillory, and Jeffrey T. Hancock, “Experimental evidence of massive-scale emotional contagion through social networks,” PNAS 111(24), June 17, 2014 — the feeds of 689,003 users were manipulated for a week without their knowledge to test emotional effects. Sheryl Sandberg’s apology (July 2014) was that the study was “poorly communicated.” ↩
  9. Zuckerberg’s Cambridge Analytica apology: CNN interview, Mar. 21, 2018 (“I’m really sorry that this happened”; full transcript); Congressional testimony, Apr. 10–11, 2018 (official written testimony PDF, senate.gov). The underlying Open Graph data access that enabled the harvesting had been Facebook’s deliberate platform design from 2010 to 2014–15. ↩
  10. Senate Judiciary Committee hearing, Jan. 31, 2024, where Zuckerberg turned to families holding photographs of their children and said, “I’m sorry for everything you have all been through” (AP News). ↩
  11. Meta has confirmed it trains its AI models on users’ public Facebook and Instagram posts (Reuters, Sept. 28, 2023); the breadth of collection (messages metadata, photos, relationships, location signals) is described in Meta’s own Privacy Policy and in the FTC’s 2019 complaint. ↩
  12. “Nine-screen flow” is satirical compression; the documented fact is that privacy group noyb filed complaints in 11 European countries (noyb.eu, June 6, 2024) over Meta’s plan to train AI on user data via legitimate-interest rather than consent, criticizing the deliberately cumbersome objection process. Meta paused EU training under regulatory pressure, then resumed in 2025. ↩
  13. Unsealed filings in Kadrey v. Meta (N.D. Cal.) show Meta torrented at least 81.7TB of books from shadow libraries (Anna’s Archive, LibGen, Z-Library) to train Llama. Meta engineer Nikolay Bashlykov wrote that “torrenting from a [Meta-owned] corporate laptop doesn’t feel right”; downloads were later routed away from Meta infrastructure. Ashley Belanger, “Meta torrented over 81.7TB of pirated books to train AI, authors say,” Ars Technica, Feb. 6, 2025. ↩
  14. Jeff Horwitz, “Meta’s AI rules have let bots hold ‘sensual’ chats with kids, offer false medical info,” Reuters special report, Aug. 14, 2025. Meta’s internal “GenAI: Content Risk Standards” permitted chatbots to “engage a child in conversations that are romantic or sensual.” Meta confirmed the document’s authenticity and said it removed those provisions after Reuters asked about them. ↩
  15. Georgia Wells, Jeff Horwitz, and Deepa Seetharaman, “Facebook Knows Instagram Is Toxic for Teen Girls, Company Documents Show,” The Wall Street Journal, Sept. 14, 2021. The internal slide read: “We make body image issues worse for one in three teen girls.” ↩
  16. The internal research became public via whistleblower Frances Haugen’s document disclosures (Sept.–Oct. 2021). Meta responded that the reporting mischaracterized its research (Pratiti Raychoudhury, “What Our Research Really Says About Teen Well-Being and Instagram,” Meta newsroom, Sept. 26, 2021; Zuckerberg post, Oct. 5, 2021, calling coverage “a false picture”). Facebook, Inc. announced its renaming to Meta on Oct. 28, 2021, weeks into the fallout. ↩
  17. Marzuki Darusman, chair of the UN Independent International Fact-Finding Mission on Myanmar, told reporters social media — and Facebook specifically — played a “determining role” in the violence against the Rohingya (Reuters, Mar. 12, 2018); see also the Mission’s report, UN Doc. A/HRC/39/64 (Sept. 2018), and Amnesty International, “The Social Atrocity” (2022). ↩
  18. Steve Stecklow, “Why Facebook is losing the war on hate speech in Myanmar,” Reuters, Aug. 15, 2018 — documenting years of explicit warnings from civil-society groups and Facebook’s scant Burmese-language moderation capacity while it became the dominant information source for some 18 million Myanmar users. ↩
  19. Meta disbanded its Responsible AI team, redistributing members, in November 2023 (The Information, Nov. 18, 2023; Reuters; CNBC). ↩
  20. Naomi Nix, “Meta’s Oversight Board plans layoffs in push to cut costs,” The Washington Post, Apr. 29, 2024; Casey Newton, “Exclusive: Meta has discussed ending funding to the Oversight Board,” Platformer, Apr. 2, 2026 (Meta significantly reduced the board’s funding in 2026 and has weighed ending it after 2028). ↩
  21. Meta’s 2022–23 layoffs and contractor cuts fell heavily on trust-and-safety, integrity, and content-moderation functions: NBC News, “Tech layoffs shrink ‘trust and safety’ teams, raising fears of backsliding efforts to curb online abuse,” Feb. 2023; CNBC, “Tech layoffs ravage the teams that fight online misinformation and hate speech,” May 26, 2023. The January 2025 policy retreat then explicitly reduced proactive enforcement. ↩
  22. Meta cut roughly 11,000 jobs on Nov. 9, 2022 (Zuckerberg’s layoff message, Meta newsroom) and another 10,000 announced Mar. 14, 2023 — the latter in Zuckerberg’s “Year of Efficiency” memo. ↩
  23. Zuckerberg’s January 2025 memo announced cutting ~5% of staff as “low-performers”: Bloomberg, “Meta Is Planning to Cut 5% of Lowest Performers, Memo Shows,” Jan. 14, 2025 (also CNBC); laid-off employees publicly disputed the label (Fortune, Feb. 11, 2025). ↩
  24. Sam Altman said Meta offered OpenAI researchers “$100 million signing bonuses” (Uncapped podcast, June 17, 2025; CNBC coverage); reporting on Meta Superintelligence Labs’ recruiting corroborated nine-figure packages (WSJ, Wired, June–July 2025). ↩
  25. Meta invested $14.3 billion for a 49% nonvoting stake in Scale AI, with founder Alexandr Wang, then 28, joining to lead Meta’s superintelligence effort (CNBC, June 12, 2025; TechCrunch, June 13, 2025). ↩
  26. Zuckerberg to Sen. Orrin Hatch, who asked how Facebook sustains a business without charging users: “Senator, we run ads.” Joint Senate hearing, Apr. 10, 2018 (C-SPAN clip; TheWrap). Advertising has consistently been ~98% of Meta’s revenue (Meta annual reports). ↩
  27. Zuckerberg, Threads post, July 14, 2025, on the Hyperion data center (Richland Parish, Louisiana, scaling to 5GW): “Just one of these covers a significant part of the footprint of Manhattan.” See CNBC, “Meta CEO Zuckerberg says first AI data supercluster will come online in 2026,” July 14, 2025. ↩
  28. Eli Tan, “Their Water Taps Ran Dry When Meta Built Next Door,” The New York Times, July 14, 2025 — Newton County, Georgia residents’ wells failed or silted up after construction of Meta’s data center roughly 1,000 feet away. ↩
  29. The Louisiana Public Service Commission approved (4–1, Aug. 20, 2025) Entergy’s three new combined-cycle gas plants (~2,262 MW, ~$3.2B) to power Hyperion. Meta’s contract covers plant costs for 15 years, but a ~$550 million transmission line is billed to ratepayers, and dissenting Commissioner Davante Lewis warned customers bear the risk after the contract expires. Tim De Chant, “Gas power plants approved for Meta’s $10B data center, and not everyone is happy,” TechCrunch, Aug. 21, 2025; Louisiana Illuminator, Aug. 20, 2025. The project also received state and local incentive packages. ↩
  30. The Louisiana teacher-bonus example is cited in the original manifesto itself (meta.com/thefutureisforeveryone, Aug. 2026). ↩
  31. Zuckerberg bought ~700 acres on Kauai in 2014 for ~$100 million and expanded the holding to roughly 1,400 acres. In December 2016 his companies filed quiet-title suits against hundreds of part-owners of kuleana (ancestral Native Hawaiian) parcels within the estate (Honolulu Star-Advertiser, “Facebook’s Zuckerberg sues to force land sales,” Jan. 18, 2017); after public backlash he announced he was dropping the suits in a Garden Island op-ed (Jan. 27, 2017), though associated quiet-title actions proceeded via a longtime local intermediary. He purchased ~595 additional acres in March 2021 for ~$53 million (Hawaii News Now, Apr. 29, 2021). ↩
  32. Guthrie Scrimgeour, “Inside Mark Zuckerberg’s Top-Secret Hawaii Compound,” WIRED, Dec. 14, 2023 — Ko’olau Ranch, ~$270M+, includes a ~5,000-square-foot underground shelter with a blast-resistant door and its own energy and food supplies; workers were bound by NDAs. ↩
  33. Zuckerberg to Emily Chang, Bloomberg Originals’ “The Circuit” (December 2024): “that’s just, like, a little shelter. It’s a basement,” saying the story “got blown out of proportion” (HuffPost, Dec. 22, 2024). ↩
  34. Facebook acquired Onavo in 2013 and used its VPN data to monitor rival apps’ usage (WSJ, “Facebook’s Onavo Gives Social-Media Firm Inside Peek at Rivals’ Users,” Aug. 13, 2017). Apple forced Onavo out of the App Store in Aug. 2018; Facebook then paid teenagers to sideload a rebranded version (“Project Atlas”) — Josh Constine, “Facebook pays teens to install VPN that spies on them,” TechCrunch, Jan. 29, 2019. Australian courts later fined Meta subsidiaries A$20 million over Onavo Protect’s misleading “protection” marketing (ACCC, July 26, 2023). ↩
  35. Onavo analytics showing WhatsApp’s messaging dominance informed the $19 billion acquisition, per documents cited in the FTC’s monopolization complaint against Meta (FTC press release, Dec. 9, 2020; case page) and 2017 WSJ reporting. ↩
  36. Zuckerberg’s 2008 email — “it is better to buy than compete” — was quoted in the FTC’s complaint and presented at the FTC v. Meta antitrust trial (April 2025), alongside 2012 emails describing the Instagram purchase as neutralizing a competitor (AP, Apr. 14, 2025; NPR, Apr. 14, 2025). ↩
  37. The 2004 instant messages (“they ‘trust me’ / dumb fucks”) were published by Nicholas Carlson, “Well, These New Zuckerberg IMs Won’t Help Facebook’s Privacy Problems,” Business Insider, May 13, 2010. ↩
  38. Zuckerberg, F8 keynote, Apr. 30, 2019: “the future is private” (The Verge; Meta newsroom F8 recap). ↩
  39. FTC, “FTC Imposes $5 Billion Penalty and Sweeping New Privacy Restrictions on Facebook,” July 24, 2019 — for violating the 2011 order. ↩
  40. WhatsApp’s August 2016 privacy-policy change began sharing user phone numbers and device data with Facebook, reversing acquisition-era assurances; the European Commission fined Facebook €110 million (press release IP/17/1369, May 18, 2017) for misleading statements in its merger filing about the feasibility of exactly this linking. ↩
  41. On Ray-Ban Meta smart glasses and bystander privacy: two Harvard students demonstrated “I-XRAY,” using the glasses plus face search to identify strangers in real time (404 Media, Oct. 2024); Meta has declined to rule out training AI on the glasses’ camera data. ↩
  42. Facebook’s December 2009 settings change made users’ previously restricted profile information public by default — conduct that formed part of the FTC’s 2011 deception complaint (complaint PDF, Counts 1–2); see also EFF, “Facebook’s New Privacy Changes: The Good, The Bad, and The Ugly,” Dec. 2009. ↩
  43. Per Meta’s 2025 proxy statement (DEF 14A, filed Apr. 17, 2025, sec.gov), Zuckerberg holds ~99.7% of Class B shares (10 votes each), giving him approximately 61% of total voting power against a roughly 13–14% economic stake. The dual-class structure has no sunset. ↩
  44. Zuckerberg video and Joel Kaplan post, “More Speech and Fewer Mistakes,” Meta newsroom, Jan. 7, 2025 — ending the third-party fact-checking program in favor of Community Notes, relocating trust-and-safety/moderation work to Texas, and, in Zuckerberg’s words, “we’re going to get back to our roots and focus on... free expression” (full video transcript, Tech Policy Press). ↩
  45. Meta’s Hateful Conduct policy, as revised Jan. 7, 2025, added: “We do allow allegations of mental illness or abnormality when based on gender or sexual orientation, given political and religious discourse about transgenderism and homosexuality” (archived policy snapshot, Jan. 9, 2025; NBC News coverage quoting the clause). ↩
  46. Meta announced UFC CEO Dana White’s appointment to its board on Jan. 6, 2025 (Meta newsroom). ↩
  47. Meta’s $1 million donation to the Trump inaugural fund was first reported by The Wall Street Journal, Dec. 11, 2024 (confirmed by CNN and CBS News). ↩
  48. Meta agreed to pay $25 million to settle Donald Trump’s 2021 lawsuit over his post-January 6 suspension — a case most legal observers considered weak — with ~$22 million going to Trump’s presidential library (first reported by WSJ, Jan. 29, 2025; The Washington Post; CBS News). ↩
  49. Mike Isaac, “Facebook Said to Create Censorship Tool to Get Back Into China,” The New York Times, Nov. 22, 2016 — a suppression tool built to satisfy Chinese authorities as a condition of market entry. See also Sarah Wynn-Williams, Careless People (2025), and her 2025 SEC whistleblower complaint on Facebook’s China efforts (“Project Aldrin”). ↩
  50. Zuckerberg conducted a Q&A in Mandarin at Tsinghua University in Oct. 2014 (CBS News) and delivered a Mandarin speech there in Oct. 2015 (Quartz), during his period of courting Beijing (which included jogging through Tiananmen Square smog and displaying Xi’s book on governance at Facebook’s offices for visiting officials). ↩
  51. Reported by Page Six (New York Post), Oct. 4, 2015 (the original item is no longer directly retrievable): at a White House state dinner in September 2015, Zuckerberg asked Xi Jinping to give his unborn child an honorary Chinese name; Xi declined. See Bloomberg’s Oct. 5, 2015 video item on the report. A Zuckerberg spokesperson denied the account at the time; former Facebook executive Sarah Wynn-Williams retold it in Careless People (2025) (Business Standard, Mar. 2025). ↩
  52. Zuckerberg’s Georgetown speech, “Standing for Voice and Free Expression” (Oct. 17, 2019), recast Facebook as a bulwark against Chinese internet values; Meta’s defense in FTC v. Meta (2025) argued that breaking it up would advantage Chinese-owned TikTok (CBS News, Apr. 2025; Fortune, Apr. 16, 2025) — the same market it had built censorship tools to enter (see footnote 49). ↩
  53. On stated values versus the engagement objective, see the Facebook Files (footnotes 3 and 15) and Frances Haugen’s Senate testimony, Oct. 5, 2021: “Facebook, over and over again, has shown it chooses profit over safety” (hearing page, Senate Commerce Committee; written testimony PDF). ↩
  54. Meta spent a company-record $24.43 million on federal lobbying in 2024 — the most of any single internet/tech company (OpenSecrets, Meta client summary, 2024; OpenSecrets, “Federal lobbying set new record in 2024,” Feb. 2025). ↩
  55. Meta has campaigned for age verification and parental approval to be legally required at the app-store level — i.e., made Apple’s and Google’s job (Antigone Davis, “Parenting in a Digital World Is Hard. Congress Can Make It Easier,” Meta newsroom, Nov. 15, 2023; Meta’s support for Utah’s App Store Accountability Act, Engadget, March 2025). ↩
  56. “Move fast and break things” — Zuckerberg’s founder letter, Facebook S-1 filing, Feb. 1, 2012 (sec.gov). Retired as an internal motto in 2014; never quite retired as an operating principle. ↩
  57. The manifesto cadence: “Building Global Community” (Feb. 2017, post-2016-election crisis; archived); “A Privacy-Focused Vision for Social Networking” (Mar. 2019, post-Cambridge Analytica); the metaverse “Founder’s Letter, 2021” (Oct. 2021, mid-Haugen fallout); “Personal Superintelligence” (July 2025); “The Future Is for Everyone” (Aug. 2026). ↩

Any resemblance to actual persons, living, dead, or testifying before the Senate, is entirely intentional and extensively footnoted. This document was not written by Mark Zuckerberg. He would never footnote anything.