Photo: stevenbartlett.com
Steven Bartlett
Host of Diary of a CEO YouTube channel.
33 checked claims
across 1 source
silver frame — 20+ claims checked
What Steven Bartlett talks about
A sample that mirrors their checked record and spreads across their subjects — not their worst claimsEvery row is reachable by scrolling its column · search matches claim text, verdicts, people, topics and sources · hover a row to trace its chain, click one to pin its detail
Claims by source
No claims match the selected verdicts.
The Man Who Calls BS On AI: They’re LYING About AI, 2027 Is When It All Breaks! | Ed Zitron
33 attributed claims Published August 2026
-
Mark Zuckerberg said that Meta would continue investing aggressively in infrastructure to meet AI demand.
A transcript of Meta’s Q2 2026 earnings call records Zuckerberg saying that as AI usage ramps, Meta would continue to invest aggressively in infrastructure to meet demand. The wording in the video is a shortened but substantively accurate attribution.
Sources used for this check
-
Eighty-eight percent of organizations regularly use AI in at least one business function.
McKinsey's 2025 global survey reports that 88 percent of respondents say their organizations regularly use AI in at least one business function. The transcript's wording is a close restatement of that statistic.
Sources used for this check
-
NVIDIA makes chips.
NVIDIA develops and sells GPUs and other chips for computing and AI systems. Although much of the physical fabrication is performed by outside foundries, describing NVIDIA as a chip maker is standard and substantively accurate.
Sources used for this check
-
The speaker's Claude would know the speaker's dog's name when asked.
This is a claim about the speaker's particular Claude account and whether the dog's name had been retained or was available in its context. Anthropic confirms that Claude can retain personal context, but the account-specific fact cannot be independently verified from public evidence.
-
Claude has memory of the past.
Claude products currently provide persistent memory and chat-search features that allow information from prior conversations to be carried into later conversations. This is product-level memory rather than necessarily human-like autobiographical memory.
-
ChatGPT had 100 million active users within the first 60 days after launch.
ChatGPT was widely reported as reaching about 100 million monthly active users roughly two months after its November 2022 launch. The transcript presents this estimate as an unqualified count of active users and as occurring within exactly 60 days.
Sources used for this check
-
TikTok took nine months to reach 100 million users.
Contemporary reporting comparing early platform growth states that TikTok reached 100 million users in nine months. In the surrounding comparison, the metric is understood as the same 100-million-user benchmark used for ChatGPT.
Sources used for this check
-
Instagram reached the referenced user scale in about 2.5 years.
A documented comparison of digital-service adoption says Instagram took two-and-a-half years to attain 100 million users.
Sources used for this check
-
The World Wide Web took roughly seven years to reach the referenced user scale.
The same comparison states that the World Wide Web took about seven years to reach 100 million users.
Sources used for this check
-
More than 60% of U.S. adults were integrated into AI tools in daily and regular routines within three years of launch, while the relevant population milestone was 40%.
St. Louis Fed research reports 54.6% overall generative-AI adoption about three years after ChatGPT's release, and its earlier study reported nearly 40% adoption after about two years. The transcript conflates different measures and time points, and its “over 60%” daily-routine figure is not supported by those data.
Sources used for this check
-
The internet took about five years and personal computers nearly twelve years to reach the comparable adoption milestone.
The St. Louis Fed's comparison says internet adoption reached a similar rate after about five years and PC adoption after about twelve years.
Sources used for this check
-
Nobel Prize-winning economist Paul Krugman said that by about 2005 the internet’s economic impact would be no greater than that of the fax machine.
The quotation is widely documented as a 1998 statement by Paul Krugman, and the Nobel Foundation confirms that Krugman received the 2008 Nobel Prize in Economic Sciences. The transcript accurately attributes the prediction to him, although the prediction itself was not fulfilled.
Sources used for this check
-
Astrophysicist Clifford Stoll wrote in Newsweek in 1995 that online databases would not replace daily newspapers and CD-ROMs would not replace competent teachers.
Clifford Stoll published the skeptical 1995 Newsweek essay “Why the Web Won’t Be Nirvana,” which included these predictions. He is commonly described as an astronomer or astrophysicist, and the quoted wording is substantively accurate despite the transcript’s “CDROM” spelling.
Sources used for this check
-
An economist named Krueger said that internet growth would slow because most people had nothing to say to each other.
The quoted prediction is attributed in the available documentation to economist Paul Krugman, not to an economist named Krueger. Krugman was also the Nobel laureate referred to earlier in the transcript.
Sources used for this check
-
All types of AI require GPUs.
AI workloads can run on CPUs or other accelerators, not only GPUs. Google’s TensorFlow Lite documentation shows an interpreter falling back to CPU execution when an accelerator delegate is unavailable.
Sources used for this check
-
AWS went down multiple times because of AI coding tools.
Reports describe at least two AWS-related incidents involving Amazon’s AI coding tools, including a 13-hour interruption involving Kiro. Amazon disputes the causal framing that the AI itself caused the outages, attributing them to user error and access-control problems.
-
The code is becoming buggier.
The claim does not define which code, whose code, or the comparison period and metric for 'buggier.' Available studies show that AI-generated code can introduce defects, but they do not establish the broader trend stated here for code generally.
-
The volume of AI activity is causing infrastructure to crash.
The claim does not identify the infrastructure, incidents, dates, or evidence connecting crashes to AI activity. Without those details, the literal causal assertion cannot be confirmed or refuted.
Sources used for this check
-
Waymo's driverless vehicles had a 68% lower police-reportable crash-involvement rate per vehicle mile than human drivers in the studied cities.
The 68% figure is supported for the specific Waymo fleet and study conditions, but the wording presents it as a general rate for autonomous vehicles and calls it an overall crash-involvement rate. The underlying study used a narrower crash definition and a limited set of Waymo deployments.
Sources used for this check
-
Waymo autonomous vehicles experienced approximately 2.1 police-reported crashes per million miles versus 4.68 for the human benchmark, a 55% reduction.
The numerical comparison and 55% calculation match the published Waymo study. However, the transcript generalizes a Waymo-specific result to autonomous vehicles broadly without stating the limited cities, period, fleet, and benchmark methodology.
-
Waymo autonomous vehicles showed an approximately 80% to 81% reduction in injury crashes compared with human drivers.
The reported range is consistent with published findings: one Waymo study found an 80% reduction in any-injury-reported crashes, while the IIHS study found 81% fewer injury crashes. The wording misleadingly generalizes those results to autonomous vehicles as a whole.
-
A person riding in an autonomous vehicle is 85% less likely to be involved in a single-vehicle crash than when driven by a human.
The 85% figure is directly reported by IIHS for Waymo vehicles compared with human drivers. The statement is misleading because it presents a limited Waymo result as a general property of autonomous vehicles and does not state the study's location and measurement limits.
Sources used for this check
-
AI agents could theoretically be more intelligent, faster, and broader in operation than a human hacker.
The report supports comparisons involving greater speed, scale, and coordination than human attackers, but "higher intelligence" and "wider" have no agreed operational definition here, and no specific human-hacker benchmark is given.
Sources used for this check
-
The AI industry is creating enormous economic growth.
The claim is not well-posed because “enormous” has no agreed threshold and the relevant time period and measure of growth are unspecified. The BEA says AI’s contribution to economic growth is difficult to measure, while the IMF reports both substantial AI-related investment and potential productivity gains.
-
AI will replace all human jobs.
The universal claim is contradicted by the ILO’s assessment that most jobs will be transformed rather than eliminated. The ILO also reports that large-scale job displacement remains limited, although some occupations and tasks face substantial exposure.
-
There was substantial hype during the dot-com era.
Historical accounts describe the dot-com period as a frenzy characterized by widespread excitement, speculative business formation, and hype.
Sources used for this check
-
Mark Zuckerberg said that an AI breakthrough produced 15 basis points of increased retention and involved processing everything posted on social media for full context.
The official earnings transcript contains related claims but not the statement as presented: it reports a 15-basis-point increase in Instagram sessions and says every public Instagram Reels and Feed post was processed through an LLM. The transcript therefore contradicts both the retention metric and the broader attribution to anything posted on social media.
Sources used for this check
-
The speaker has conducted almost 700 interviews.
The statement is a personal estimate, but no reliable independent source establishes the number of interviews conducted. Public podcast and promotional material confirm the claim was made, not that the approximate count is accurate.
Sources used for this check
-
The conversation cards contain questions taken from The Diary of a CEO and identify the person who wrote each question.
The product's official description says each question is taken directly from The Diary of a CEO and describes the cards as featuring questions from thought leaders and experts. This supports the claim that the cards use guest questions and identify their authors.
Sources used for this check
-
The recommendation algorithm identifies the viewer as someone who watches the show but has not subscribed.
The claim concerns a personalized algorithmic classification, but it does not identify the platform or provide access to the viewer-specific data needed to verify it.
-
Sundar Pichai said the risk of underinvesting was dramatically greater than the risk of overinvesting.
Pichai made this comparison publicly while discussing Google's AI investment strategy. The transcript's wording matches the reported quotation.
Sources used for this check
-
Andy Jassy said Amazon was investing approximately $200 billion in 2026 capital expenditures and not doing so on a hunch.
Jassy used this wording in Amazon's 2026 shareholder communication. Amazon later raised its 2026 capital-expenditure expectation, but that does not make the quoted earlier statement false.
-
Mark Zuckerberg said Meta would continue investing aggressively in infrastructure to meet demand and preferred building capacity before it was needed rather than being late.
Meta's prepared remarks contain both the stated preference for building capacity ahead of demand and the rationale that infrastructure projects have long lead times. The transcript is a shortened paraphrase of Zuckerberg's wording.
Sources used for this check