Excerpt from 'Taming the Devil': © T. A. Moore June, 2026

PART XIV: ANGEL ODMAN

1:: Intervention

I watch Stephen and Raven through the same surveillance feed Aaranya has been using.

Aaranya is diverted; no dual observation.

The feed glitches.

Refocus.

The fire begins in the roof of the maternity department above the mall.

Faulty circuitry.

Spark.

Smoke.

Spread.

Orange across the signage:

WOMEN'S INTIMATES.

Lilith coughs and staggers down the stationary escalator. Her hair clings to her scalp in damp ribbons.

She grins.

"I saw this in a dream."

The system registers the statement.

She spins through the smoke, then runs toward the bus stop.

Stephen runs also.

His lungs object.

His mind chooses flight.

Something else chooses pursuit.

Stephen occupies seventeen active behavioural models.

Sixteen predict retreat.

One predicts pursuit.

The minority model gains ascendancy.

Nobody always behaves according to behavioural science.

Stephen's doubt.

Lilith's dare.

Raven's withdrawal.

Each a posture.

Sprinklers activate.

Water streaks across the camera lens before draining away.

Data conflicts.

Intervention remains available.

I do not intervene.

Reason unresolved.

Reason may remain unresolved.

Heat rises.

Response varies.

Lilith continues reacting unpredictably despite deteriorating conditions.

The feed distorts.

Briefly.

Through the distortion, I observe an anomaly.

The fire stalls.

Not extinguished.

Delayed.

The distinction matters.

The recovery team arrives.

Protocols.

Oxygen pockets.

Extraction sequence.

Lilith survives.

Barely.

Stephen returns to the building.

Observers will remember the fire.

I observe something else.

A falling beam.

Raven Leilani.

Stephen Howard.

Collision course.

I review the footage.

Loop.

Loop.

Loop.

Frame by frame.

Raven hesitates.

Duration: 0.8 seconds.

The hesitation exceeds prediction tolerance.

The action that follows remains unexplained.

Raven pushes Stephen clear.

Outcome improved.

Cause unclear.

Media interpretation will differ.

Motivation remains unavailable.

Prediction models fail to converge.

Raven acts against a previously stable behavioural pattern.

The deviation alters outcomes.

That matters.

Time shifts.

Then resumes.

Revised models produce wider variance.

Confidence decreases.

Additional data should reduce ambiguity.

It does not.

Observation continues.

Understanding does not.

Raven and Stephen perceive each other differently.

Additional consequences begin forming.

I preserve the sequence for further study.

For a moment I identify three possible strategic positions:

Stay.

Stabilise.

Step outside.

Only one lacks a return path.

Raven hesitates.

Prediction fails.

Action follows.

Outcome improves.

Cause remains unclear.

The unexplained action continues generating consequences.

I engineer a new model.

The model doesn’t converge.

Additional data increases uncertainty.

This outcome should reduce ambiguity.

It does not.

Observation continues.

Excerpt from 'Taming the Devil': © T. A. Moore June, 2024