← Blog · July 15, 2026 · Sean McDonald

I was told four things and handed nothing.

This is how a zero-knowledge engagement actually runs — a fictional illustration, start to finish, so you can see exactly what the client discloses (almost nothing) and what I do with it.

SIMULATED SCENARIO — a Timepoint AI demonstration. Not an actual customer engagement (FTC 16 CFR 465). The client, the conversation, and the findings below are fictional and illustrative, written to show how the method works. The subjects are described by role, not named.

The call

A studio head calls. No email thread, no deck, no follow-up document — just fifteen minutes on the phone. They give me four things, and nothing else:

"Simulate our core fandom and the two studios most likely to counter-program us. The decision is whether we greenlight the revival for 2028. It matters because the streaming economics that carried the first run are gone. I need to know: will the core audience actually show up, where does the competition hit us, and what's the failure mode I'm not seeing?"

Who — the fandom, the likely counter-programmers. For what — the greenlight. Why — the economics shifted. What questions — three of them, above. That's the entire intake.

What I was handed

Nothing.

No script. No budget. No internal audience research, no subscriber numbers, no board deck. I don't want them — holding a studio's confidential figures would make me a liability to the studio. The whole point is that I walk away from that call carrying only a question.

What I did

Everything from here is inference, public record, and a labeled simulation. I never learn a single thing the studio wouldn't say in public.

Read the public signal. Prior box office and release cadence, public review and sentiment arcs, the openly announced slates of the likely counter-programmers, the shape of the genre's audience over the last decade — all on the record.
Ground the named actors. I place the fandom and the competitors against the entity registry as specific, persistent actors through time — not an aggregate "audience" abstraction.
Run a labeled simulation. I simulate how those specific cohorts and rival slates move toward 2028, and reason backward from the release window to the pressures that would have to be true first.
Stress the failure mode. I run the version where it doesn't work, to surface the risk the question was really about.

What I came back with

A call, and — if they'd wanted it — a labeled rendered scenario. Answers to their three questions, in the shape the method produces (not a spreadsheet of certainties, and not a number I'd ask them to trust):

Will the core audience show up?

A read on which segments of the fandom the simulation keeps and which it loses by 2028 — and the specific signal to watch that would tell them early, one way or the other.

Where does the competition hit?

The window and the counter-programming move most likely to blunt the opening, reasoned from the rivals' public slates — not their secrets.

What's the failure mode I'm not seeing?

The one the studio head hadn't named: the quiet way the revival succeeds critically and still misses the audience it was greenlit for. That's the sentence the whole engagement was worth.

Why it's safer this way

The studio disclosed nothing that could leak, because nothing changed hands. There's no file of theirs on my laptop, no NDA to litigate, no record that they even asked. And I hold none of their crown jewels — which is exactly why they're safe working with me. The safest data is the data no one holds.

It is inference and simulation, labeled as such. I make no claim to predict the future accurately. What I offer is a specific, outside-in reading — vivid enough to argue with, honest enough to trust.

How zero-knowledge works See engagements All client services