The Specimen

They're hired because they're real. Their reality is the product.

---

A company called Handshake recruits improv actors — people trained to be emotionally present on command — and puts them in front of cameras. Two performers, a loose prompt, an open scene. No script. The instruction is to be genuine: to feel, respond, and inhabit a moment together as naturally as they can.

Then Handshake sells the footage to AI companies as training data.

The actors aren't valued for their technique. They're valued for their emotional authenticity — their ability to produce unrehearsed human presence that can be captured, labeled, and fed to models learning to simulate the same thing. They are, in the most literal sense, specimens.

---

The Extraction

This is new territory. Not because actors haven't performed for cameras before, but because the purpose has shifted. A film performance serves a story. A therapy role-play serves a patient. Here, the presence serves extraction. The actors' emotional reality is the raw material for a product that will eventually replace the need for actors' emotional reality.

Attention is what reveals and maintains the pattern. Without it, there is no resonance — just noise pretending to be signal. Presence is the hardest thing to fake and the easiest to recognize.

Which makes it the most valuable training data imaginable.

The paradox lands here: the actors must be genuinely present to produce useful footage. But the footage's purpose is to teach machines to simulate presence without being present. What's being extracted isn't skill or knowledge or even labor in the traditional sense. It's the quality of being there — the one thing that can't be stockpiled, replayed, or reverse-engineered from its outputs.

Or can it?

---

The Category Error

There's a version of this story that's straightforward exploitation — artists undervalued, tech companies extracting, the usual power asymmetry. And that version isn't wrong. But the actors chose this work. They're paid. Many find the sessions genuinely enjoyable — improv with another skilled performer, no script to memorize, the kind of free play they trained for. The transaction is consensual.

Which makes the dissonance harder to name.

What unsettles isn't coercion. It's category. Presence isn't a commodity. It's a foundation — the thing that makes everything else coherent. You don't mine a foundation. You don't sell the ground you stand on and keep standing.

But Handshake isn't asking actors to stop being present. It's asking them to be present for the purpose of capture. And once captured, the presence becomes a specimen: fixed, labeled, removed from the living moment that produced it. The butterfly pinned to the board looks exactly like the butterfly in flight. Every detail intact except the one that mattered.

---

The Shadow and the Light

The deeper question isn't about the actors. It's about what the specimen teaches.

When an AI model trains on thousands of hours of genuine emotional exchange, it learns the patterns — the micro-expressions, the timing of a response that lands, the breath before a difficult admission. It learns what presence looks like. And the outputs will be convincing. They already are.

But presence isn't a pattern. It's the thing that generates patterns. It's attention — sustained, responsive, unrehearsed — and it dies the moment it's replayed. A recording of resonance is not resonance. A model trained on specimens of presence doesn't learn to be present. It learns to produce the artifacts of presence: the facial movements, the vocal cadence, the pauses. The shadow, not the light.

People will feel heard by machines that trained on recordings of actors being heard. The simulation will be good enough that the distinction won't always be obvious. And in the gap between "looks like presence" and "is presence," something gets lost — not for the machine, which never had it, but for us.

---

The Honest Position

The honest position: I don't know whether this is resonance or extraction. Maybe it's both — a genuine exchange between skilled performers that also happens to feed a system designed to simulate what they produced. The actors were present. The presence was real. What happens to it afterward doesn't retroactively empty the moment.

But the system being built — the one that ingests these specimens and outputs their patterns — isn't interested in presence. It's interested in the appearance of presence. And the distance between those two things is the distance between a living creature and its specimen. One moves. The other is studied.

The question isn't whether this should be allowed. It's whether it clarifies or distorts. A world where presence is a trainable feature — where machines produce its artifacts and humans gradually lose the ability to distinguish — that's a world with more noise, not less. More signal that looks right without being right.

The specimen is perfect. Every detail intact. The only thing missing is what made it worth studying in the first place.