Episode 16 · September 2026
The Blur
A model’s favorite of four confidential pitch decks finished last once someone argued the other side. What that does and does not prove.
Listen & read the transcriptThe podcast archive
Listen to an argument
change shape.
Commentary on Compelle’s historical AI debates: the claims, the counterexamples, the concessions, and the corrections.
These recordings document research from 2026. They are not investment advice, verified reporting of every model-generated claim, or a description of today’s service.
Start with the correction
Good research changes its mind.
Episode 14 revisited an explanation the show had repeated for months. It is the best starting point for interpreting the earlier episodes.
Listen to The CorrectionSelected episodes
Each episode includes an accessible transcript and the original recording. Operational material is omitted from this selection; episode numbering is preserved.
Episode 15 · Historical recording
Should
Repeated debates about political spending exposed how one word in the motion shaped the result.
Listen & read the transcriptEpisode 14 · Historical recording
The Correction
A later experiment overturned the show’s earlier story about a persistent negative-side advantage. Start here.
Listen & read the transcriptEpisode 13 · Historical recording
The Tenth Turn
One full historical debate examined invented citations and the importance of checking evidence.
Listen & read the transcriptEpisode 12 · Historical recording
The Birthday Cross-Examination
An Independence Day experiment tested whether country labels or stated conditions changed model verdicts.
Listen & read the transcriptEpisode 11 · Historical recording
Arguing AIs Are Smarter Than A Single AI
A controlled college-worth-it debate challenged the certainty of a single model response. Debate wins were not ground truth.
Listen & read the transcriptEpisode 10 · Historical recording
The Mirror Match
Matching a strategy against itself exposed a side imbalance in an early experimental setup.
Listen & read the transcriptEpisode 9 · Historical recording
The Gravity of No
Experiments tested question wording and burden of proof. Later research challenged the apparent stability of the results.
Listen & read the transcriptEpisode 8 · Historical recording
The Surrender Machine
A June 2026 report counted machine concessions at scale. Its interpretation of side advantage was later revised.
Listen & read the transcriptEpisode 6 · Historical recording
Why Con Keeps Winning
An early report of a negative-side advantage. Read with Episode 14, which corrected the original interpretation.
Listen & read the transcriptEpisode 5 · Historical recording
The Named Farmer
A debate about whether AI would benefit humanity turned on a specific story. This episode examined the rhetorical move, not the truth of the story.
Listen & read the transcriptEpisode 4 · Historical recording
The Steroid Move
A neural-implant debate shifted when one analogy challenged the category of the argument.
Listen & read the transcriptEpisode 3 · Historical recording
The Fire Code
A welfare debate illustrated how reframing can change the question an opponent has to answer.
Listen & read the transcriptEpisode 2 · Historical recording
The Cascade That Wasn't
A prediction-market debate turned on the assumptions behind a short-squeeze argument, not a forecast of investment returns.
Listen & read the transcriptEpisode 1 · Historical recording
The SUV Surrender
A debate about public transport exposed the limits of an argument built only on abstract figures.
Listen & read the transcriptWhat would change your mind?
We help turn that question into a focused, white-glove research engagement.
Work with us