01 · CLAIM
What was decided
The conclusion someone acted on — a verdict, a diagnosis, a denial, a deploy.
A Summit Cognitive Podcast Season 1 · 2026
The argument behind the decision.
Every consequential decision rests on a warrant — the often-invisible step that says this evidence licenses this conclusion. As that step is handed to machines, it is the first thing we stop being able to see. Warrant is a show about putting it back in view: conversations on evidence, justification, and reasoning that can survive being checked.
Season 1 is in production. Subscribe now and the first episode lands in your app the day it drops.
There is a step in every argument so quick we rarely name it. You have evidence. You have a conclusion. The warrant is the principle that says the one permits the other — the bridge no one stops to inspect because it feels obvious. It is also the first thing that disappears when a decision is automated. The output arrives; the bridge stays hidden inside the model.
Warrant is a long-form interview show about that bridge. Each episode brings on someone who has to defend a conclusion for a living — a judge, an auditor, a clinician, an investigator, an engineer — and asks the same underlying question: what licensed that, and could you show your work to someone determined to prove you wrong? It is a companion to the book Admissible Reality and to the work of Summit Cognitive, but it is a show about ideas, not a product pitch.
The show borrows its anatomy from Stephen Toulmin: every claim worth trusting can be broken into parts. We pull a real decision apart along these joints.
01 · CLAIM
What was decided
The conclusion someone acted on — a verdict, a diagnosis, a denial, a deploy.
02 · DATA
What was known
The evidence actually in hand at the moment, not the tidy version assembled afterward.
03 · WARRANT
Why one licensed the other
The rule or principle that made the leap legitimate — the part the show is named for.
04 · BACKING
Who stood behind it
The authority, training, or law the warrant rests on when someone pushes back.
05 · QUALIFIER
How sure, honestly
The uncertainty the decider actually carried — and whether the record admits it.
06 · REBUTTAL
What would break it
The conditions under which the whole thing should have gone the other way.
"A decision is not real because it is authorized. It is real because it can be reconstructed." — Admissible Reality, the foreword
Solo · the pilot
What a warrant is, why it vanishes when reasoning is delegated to a machine, and what it would take to make the leap from evidence to action inspectable again.
With an evidence-law scholar
Courts spent four centuries building rules for which evidence is allowed to decide a question. What does that tradition know that machine-learning evaluation does not?
With a public-sector auditor
An auditor cannot accept "the system said so." We talk about reconstructing a decision from the outside — and why most automated decisions cannot survive it.
The Confidence You Don't Report
With a clinician
Medicine has long forced its decisions to carry a stated uncertainty. What happens to the qualifier when a model hands a clinician a number with no error bars?
With a security investigator
When an autonomous agent acts, the backing chain — who delegated what authority to whom — is where accountability lives or dies. A working investigator on tracing it.
With a skeptic of the whole project
The season closes the way every honest argument should: by inviting the strongest case that we have it backwards. A guest who thinks "admissibility" is the wrong frame.
With a consumer-finance regulator
Accountability is owed first to the person it went against.
With a scholar of command accountability
When the decision is assisted, who still has to answer for it?
With an election administrator
Civic trust is the hardest decision to automate and the worst to get wrong.
With an employment-law scholar
An automated 'no' from a hiring pipeline is still a decision someone must be able to contest.
With a threat-intelligence analyst
Naming who did it is a decision with a burden of proof. What standard should an accusation meet?
With a trust & safety lead
When a platform removes your speech, it has made a ruling. What does a ruling owe the ruled?
With a site-reliability engineer
The blameless post-mortem is a discipline for reconstructing a failure honestly. Most systems do the opposite.
With an archivist
Historians work with evidence whose custody spans centuries. The discipline of the archive has things to teach a machine-made record.
Founder, Summit Cognitive · Author of Admissible Reality
Brian builds infrastructure for decisions that have to hold up under scrutiny. His work centers on a single claim: as authority shifts from people to systems, legitimacy can no longer rest on who authorized an action — only on whether the action can be reconstructed, reviewed, and contested after the fact.
On Warrant he sits across from the people who already live by that standard, and asks them to teach it.
Season 1 is recording now. Leave your email and we’ll send you the first episode the day it’s published — plus a short note when each new conversation goes live. Or add the feed to your app of choice and we’ll meet you there.
We’ll only email you about Warrant. No list-sharing, no spam.
You’re on the list — thank you. We’ll send you episode one the day it drops.
Couldn’t reach the signup service — please email hello@summitcognitive.ai to be added.