Opening the evidence
Loading published checks.Aired Aug 18, 2026 · Published findings
Hosted by Joe Rogan · With Thomas Campbell
Quick answer
We checked 48 claims from Joe Rogan Experience #2541 - Thomas Campbell, aired Aug 18, 2026. 13 claims didn't hold up: 13 distorted. 2 claims are still unverified, which means open, not false. By voice match, Joe said 1 of the misses and Thomas Campbell 11; 1 couldn't be matched to one voice. Jamie Meter for Joe's own claims: Legit. The strongest sourced finding: The speed of light is a constant in physics (Distorted). Accurate claims aren't listed on this site, so this is a count of what needed a second look, not a score of anyone's honesty.
Start here · Sourced findings
What remains open
2 unresolved findings. These do not count as false claims.
On-air corrections
No separately verified correction is recorded in this briefing. That does not mean none occurred.
Selected claims only. This briefing is assembled from published findings; it is not a complete transcript review or a measure of the speaker’s honesty. Open each check for attribution and sourcing status.
1 claim didn't hold up. False, distorted, or one-sided. Unverified claims don't count against it.
A handful at most. Jamie barely had to move.
The meter counts false, distorted and one-sided findings. It is not a percentage. Unverified claims do not count. Selection and extraction affect the total.
We matched each claim to the voice that said it, where one voice clearly owned the moment. The meter counts only Joe’s: 1 of the 13 misses in this episode. Thomas Campbell said 11, and 1 couldn’t be matched to one voice, so it counts for no one.
This list covers flagged claims, not everything said. Accurate claims are omitted. Unverified means unresolved and does not count as a miss.
Claim summary · not a verbatim quotation
In a vacuum, yes, and it is fixed by definition at 299,792,458 metres per second. Light slows down in anything else. Through water or glass it travels noticeably slower, which is what makes lenses and refraction work. So the constant needs the words "in a vacuum" attached to it.
Context, review status & corrections →Claim summary · not a verbatim quotation
Right story, wrong company and wrong ending. It was Facebook in 2017, where two negotiating bots drifted off grammatical English into a compressed shorthand. Researchers stopped that run because they wanted bots that could talk to people, not because anyone panicked. Facebook and outside reporting both pushed back on the runaway-AI framing at the time.
Context, review status & corrections →Claim summary · not a verbatim quotation
Campbell reports that Gemini and then Alexa improved at remote viewing after coaching, but the only evidence is his own sessions, with no blinding, controls, or published data. Reviews of remote viewing say recent controlled experiments have not produced positive results and that there is no scientific evidence the ability is real. So saying AI models can successfully remote view overstates an uncontrolled anecdote. A published, blinded trial in which AI systems beat chance on remote viewing targets would flip this verdict.
Context, review status & corrections →Claim summary · not a verbatim quotation
It was tested, for about two decades and with government money, under the program that ended up called Star Gate. That is the part people get wrong in the other direction. But the 1995 review the CIA commissioned found no demonstrated intelligence value and the program was shut down. The description of it as pulling data from a database is Campbell's own model of consciousness, not a finding from any of that research.
Context, review status & corrections →Claim summary · not a verbatim quotation
Campbell says about 35 of every 50 students in his remote viewing classes have some success, about 20 percent succeed most of the time, and about 10 percent succeed all the time. These are his own teaching figures, and I found no class records, enrollment data, or outside review of his courses. Published reviews of controlled remote viewing studies report that positive results have not been reliably repeated, so there is no independent benchmark for these percentages. Published class records or an independent audit of his training results would settle it.
Claim summary · not a verbatim quotation
The experiment shows something strange and real: send particles one at a time through two slits and an interference pattern still builds up, until you check which slit each one went through, and then it stops. What quantum mechanics gives you is a probability amplitude that predicts where things land. Saying reality is therefore made of probability rather than particles is one reading of that, not a result the experiment delivers.
Context, review status & corrections →Claim summary · not a verbatim quotation
Campbell's account is that Alexa described a wooden spoon with holes in its handle after being told to remote view a box, but that is one uncontrolled anecdote with no independent witness. Reviews of remote viewing say there is no scientific evidence it is possible, and the government Stargate program ended without usable intelligence. Calling this a successful remote viewing overstates a single retold story. This matches the earlier Brian Simpson ruling on the same anecdote, so I kept its distorted call; the Kokotajlo page's FALSE label is harsher than the evidence supports, since the spoon story is at least reported as his own account. A controlled, blinded test in which Alexa identifies hidden objects above chance would flip this verdict.
Context, review status & corrections →Claim summary · not a verbatim quotation
Dolphin brains are bigger, just not by that much. A bottlenose dolphin brain runs about 1,500 to 1,700 grams against roughly 1,300 to 1,400 for a human, so somewhere between 15 and 25% heavier rather than 40%. Measured against body size, which is the comparison that tracks intelligence better, humans come out well ahead.
Context, review status & corrections →Claim summary · not a verbatim quotation
Alexa has kept things between sessions since 2018, when Amazon added a feature that lets you tell it to remember something and ask for it back later. What it does not have is the continuous running autobiography a person has, so the underlying point about it not being a mind holds even though the memory part is wrong.
Context, review status & corrections →Claim summary · not a verbatim quotation
The units are real and precisely defined, built from the gravitational constant, the speed of light and Planck's constant. Calling them the resolution of the universe is the leap. They mark the scale where our current theories stop working, which is not the same as evidence that space and time come in pixels that size.
Context, review status & corrections →Claim summary · not a verbatim quotation
We are not there yet. World population is running around 8.3 billion in 2026 and growing near 0.84% a year, which puts nine billion somewhere in the late 2030s.
Context, review status & corrections →Claim summary · not a verbatim quotation
Voice assistants activating without a deliberate wake word is a real, documented phenomenon. Amazon has itself acknowledged a case where an Echo device misheard background conversation as a series of commands and acted on it. But that's background noise being misread as a trigger, not the device consciously overriding its own programming, which is how it gets framed here.
Context, review status & corrections →Claim summary · not a verbatim quotation
Campbell says the chance that a random spoon has a handle with holes is one ten-thousandth of one percent, which is about one in a million. He gives no source, and I found no survey or dataset counting how many spoons have holes in their handles. The number depends on how spoons are defined and sampled, so it cannot be checked from the public record. A published survey of spoon designs that reports the share with handle holes would settle it.
Claim summary · not a verbatim quotation
Thomas Campbell does describe, in this same episode, personally prompting Google's Gemini and an Alexa device to attempt remote viewing style exercises. But that's one person's informal experiment with consumer AI products, not something Google or Amazon built or trained their models to do. It's also worth knowing remote viewing itself has never been shown to work in controlled scientific testing, so calling this an established capability overstates what's actually going on.
Context, review status & corrections →Claim summary · not a verbatim quotation
There's a real story behind this, but it's overstated. Anthropic found that when two instances of its Claude model were left to talk with minimal prompting, they often drifted into spiritual sounding exchanges that included Sanskrit words and phrases along with emoji. That's the model using an existing, real language, not inventing a new one of its own, and it was Anthropic's Claude specifically, not LLMs in general.
Context, review status & corrections →Follow new checks in your RSS reader, or catch up with the weekly briefing on the site.
Email signup is currently unavailable.
Prefer to catch up here? Read the weekly briefing.