A reference site is only worth trusting if it admits its own mistakes. When we find something wrong — a bad source, an outdated figure, a verdict we got wrong — we fix it and record it here, newest first. Entries are never quietly deleted. See our methodology for how verdicts and sources work, or tell us about an error through the comment form on any myth.
What changed: Merged two near-duplicate myth pages into the page that already covered the same claim. mars-timeline-broken-promises now redirects to humans-on-mars-missed-deadline, and musk-never-ships-on-time now redirects to musk-never-delivers-on-his-promises. Both pairs had the same verdict (mostly true for the Mars deadline, mostly false for "never delivers"). The surviving pages took over the retired pages’ questions that they did not already answer (reworded where the facts had moved on, for example on NASA’s Artemis landers; one Mars question was dropped because its answer was not supported by the sources) and their sources (Fortune on the 2016 Mars plan, NASA on the Artemis lander, Electrek on missed 2025 predictions, SpaceX on Falcon 9 reuse). No verdict changed.
Why: Two pages answering the same question split readers and search engines between near-identical answers. Old links still work: each retired address permanently redirects to the page that replaced it.
What changed: We had described Tesla’s Supercharger network as "the largest and most reliable DC fast-charging system in the world", or similar, in more than a dozen places, without a source. It is not the world’s largest: China’s TELD alone reported more than 580,000 DC charging terminals at the end of June 2026, against Tesla’s 84,465 Supercharger connectors. It is the largest in the United States by number of ports, with about half of all US public DC fast-charging ports (49.3% in Q3 2026, per charging-data firm Paren). We now say that, with the source. We also removed "most reliable", "so clearly superior", "widely considered the best", "on technical merit", "years ahead of demand", claims that Tesla "dominates" or "leads" on charging, and a sentence saying independent surveys rate Superchargers the most dependable network with about 99.9% uptime: we could not source the 99.9% figure or the surveys.
Why: A superlative was repeated from page to page without anyone checking it against current data. Every remaining claim about the network’s size now says what it is measured by and cites a dated source.
What changed: Refreshed our living statistics from company filings and independent trackers, and corrected errors found on the way. Errors: the homepage and dashboard had been showing figures stored in June 2026 rather than our current ones (for example a $400B SpaceX valuation after its $1.75 trillion IPO), and now show our current figures; the Tesla Supercharger count was 36,500, far too low (Tesla reports 84,465 connectors at 30 September 2026); we said 2025 was Tesla's first down year for deliveries, but 2024 was already lower than 2023; our Falcon 9 landing count was a Falcon 9-only July figure, mislabelled as including Falcon Heavy and dated June; SpaceX's 2023 and 2024 revenue did not match its IPO prospectus (they now follow it; it includes xAI and X) and 2025 growth is about 33%, not about 30%; Starlink's 2024 and 2025 subscriber year-ends now follow SpaceX's filings; xAI revenue is now described as SpaceX's AI segment, which includes X; Neuralink pages said all participants were "actively using their devices", which Neuralink's announcement does not say, and called them implanted where Neuralink says enrolled. We removed claims we could not source: that Starlink is about 75% of all active manoeuvrable satellites, that it is SpaceX's most consistently profitable segment, that SpaceX's cadence is the highest of any launch provider in history, "brand headwinds" as the cause of Tesla's sales fall, the world's-largest-charging-network line on the Supercharger statistic, and early chart points we had not checked against a filing (Starlink 2021–2022 subscribers, SpaceX 2022 revenue, Tesla 2019 deliveries). Updates: Starship integrated flights 12 to 14; record booster flights 35 to 37; Starlink satellites in orbit 10,413 to 11,204; Starlink consumer subscribers 12.0 to 13.6 million; Tesla cumulative deliveries about 9.2 to 10.2 million; Falcon 9 landings 617 to 659; a new SpaceX quarterly revenue statistic. New: a year-by-year launch table on /launches (which does not count the 2016 AMOS-6 pad explosion as a launch) and NHTSA recall and complaint data on the pages of Tesla models with NHTSA records.
Why: These figures go out of date as companies report, and a few had been carried forward without a source that supported them. We added a monthly job that re-reads SEC EDGAR, Launch Library 2 and NHTSA and proposes the changes for human review, and our automated data checks now fail if our SpaceX quarterly revenue or yearly launch counts disagree with figures from those sources.
What changed: While building new detail pages we re-checked existing text against current sources and corrected it. Facilities: SpaceX's Hawthorne plant is no longer described as its headquarters (SpaceX's 2026 SEC filings list Starbase, Texas); Colossus GPU and build-time figures are now attributed to the company; Gigafactory Nevada is described as Tesla-owned with Panasonic as a partner; Lathrop is no longer called Tesla's single Megapack site; Fremont no longer lists the Model S and X; Shanghai production start follows Tesla's 10-K (December 2019). xAI: removed an unsourced founding day, revenue figure and valuation framing; added the February 2026 SpaceX acquisition. People: updated outdated roles (including Jared Isaacman, Igor Babuschkin, David Sacks, Tom Zhu and Jared Birchall) and removed unsupported claims. Biography: Stanford says it has no record of Musk enrolling (the page had said he briefly enrolled); Penn conferred his degrees in 1997. X Money: licence count, interest tiers and launch date updated; the FinCEN registrant is X Payments LLC. DOGE: added that the temporary organisation's term ended on 4 July 2026; corrected how a GAO 27.5% figure was described. Starship Flight 12 now notes the FAA mishap investigation.
Why: Several of these entries had gone stale as events moved on, and a few stated things their sources did not support. Every new or changed sentence was written from fetched sources and checked twice by an independent reviewer before publishing.
What changed: Updated the bodies, sources and verdicts of the seven myths held at "Contested" in the October review. Factual corrections: doge-savings-were-fabricated now reports GAO's $81.1 million overstatement on lease terminations (the page had said about $59M), adds the $1.7 billion in savings DOGE kept reporting on a Defense Health Agency contract it had agreed not to cancel, notes GAO could reconcile DOGE's stated methodology for only about 27.5% of the contract savings it examined, and dates the review to the GAO report (2026-08-06); starship-rains-debris-on-people adds the March 2025 Flight 8 breakup and FAA Florida ground stops and removes an unsourced clause that SpaceX "moved to compensate affected residents"; grok-makes-nonconsensual-deepfakes now describes Grok Imagine's named "spicy mode" and the Taylor Swift deepfake it produced; mars-timeline-broken-promises adds contemporaneous 2016 and 2020 sources; grok-programmed-to-push-white-genocide was re-sourced to the post-incident xAI statement and the published Grok prompts on GitHub; xai-no-safety-guardrails replaced an unsupported "does red-teaming" line with a security firm's red-team result. We also removed or corrected wording the cited sources did not support: a "public apology" and "bipartisan letters" reference on grok-mechahitler-unsafe, a quoted "if everything goes right" phrase on mars-timeline-broken-promises, "pre-cleared ocean corridors" on the Starship page, and a "within hours" rollback claim on grok-programmed-to-push-white-genocide. Verdicts: mars-timeline-broken-promises to "Mostly true"; starship-rains-debris-on-people, doge-savings-were-fabricated and grok-makes-nonconsensual-deepfakes to "Misleading"; grok-mechahitler-unsafe, xai-no-safety-guardrails and grok-programmed-to-push-white-genocide remain "Contested".
Why: These seven were deliberately held pending firmer sourcing. On review, some sources did not carry the facts the pages asserted (a figure, a date, an unsourced compensation claim), and newer 2025–2026 reporting materially changed the picture. We corrected the facts first, then rated each verdict to the evidence rather than leaving the vague "Contested" label.
What changed: Continued the October 2026 verdict review, changing only the verdict label and the wording that states it, plus the reviewed date and the one-line summary — the sourced facts on each page were left unchanged. To "Misleading" (the facts are accurate but the sweeping conclusion is not supported): covid-zero-cases-prediction-crank, starbase-destroys-environment, sweatshop-work-culture, musk-politics-destroyed-tesla, thai-cave-pedo-comment. To "Mostly false" (the headline claim is contradicted by the record): musk-never-ships-on-time, aligning it with the near-identical musk-never-delivers-on-his-promises. To "Mostly true" (a regulator substantially bears the criticism out): tesla-autopilot-false-advertising.
Why: Continuing the move away from the vague "Contested" label so each verdict matches the reasoning its own page lays out. Several further candidates were deliberately held at "Contested" pending page updates for newer reporting or firmer sourcing — the Grok-safety entries (grok-mechahitler-unsafe, xai-no-safety-guardrails, grok-programmed-to-push-white-genocide, grok-makes-nonconsensual-deepfakes), the Mars-timeline, DOGE-savings and Starship-debris pages, and fremont-factory-injury-rate.
What changed: Re-rated ten myths from "Contested" to the verdict their own evidence text already argued. Mostly true: musk-sues-critics-to-silence-them, doge-cuts-killed-people, musk-interfering-european-democracies. Mostly false: musk-never-delivers-on-his-promises, fsd-safety-stats-fabricated. Misleading: sec-hid-twitter-stake-fraud, neuralink-failed-threads-retracted, musk-bought-2024-election, boring-company-environmental-scofflaw, eu-fined-x-for-breaking-the-law. The evidence and sources on these pages were not changed; only the verdict label and the wording that states it.
Why: An evidence review found that "Contested" was a vague label for entries whose own text concluded the claim was largely accurate, largely inaccurate, or accurate in its facts but misleading in its framing. The ratings now match the reasoning shown on each page.
What changed: Removed or replaced ~26 source URLs that an internal audit found to be fabricated or incorrect (plausible-looking but non-existent links).
Why: They failed an existence check. We added a build-time citation gate that blocks the site from publishing any source URL that cannot be verified to exist, so this cannot recur.