When the AI Summariser gave up

Out of curiosity, I fed Rage into a well-known AI summariser.

It dutifully read all 250 pages, thought for a while, and then produced a neat little mechanical summary—about six pages long—and stopped.

No conclusion. No mention of the final shift from Rage toward Edge. Just a kind of dignified silence, as if it had decided that was quite enough.

At first, I wondered if the problem was my storytelling. Maybe it really was incomprehensible—too many timelines, too many versions of the same people, too much humming recursion. But when I looked closer, the AI had handled all the straightforward bits just fine: Kyiv hotel room, stolen jet, oligarch dinner. It just couldn’t describe what happens when the narrative folds in on itself and reality starts to shimmer.

It could handle the surface descriptions, but not the porous permeability of internal thought. 

An example from a tense and visceral scene about a third of the way through the novel: 

The Oligarchs’ Dinner and Power Struggles

The narrative describes a tense oligarch dinner where power dynamics and survival instincts are at play. Vitalievich, an oligarch, hosts a meeting to address the economic crisis and the threat posed by Putin’s regime. The discussion reveals the oligarchs’ fear of losing power and their desperation to maintain their wealth amid economic decline.

• 
  

  The meeting is interrupted by an assassination attempt, highlighting the treacherous nature of their political landscape.

  
  

  When the story later folded into recursion, the analyser fell silent. That silence is the break-point: the moment where linear intelligence meets nonlinear art.

We can see the AI has flattened the whole scene. No smoking guns, no leather-clad ‘heavies’. No tension.

The AI interpretation sought sequence bypassing the resonance offered by Rage. It wanted closure; the novel demanded surrender.

In its confusion, the machine performed the book’s argument perfectly. It became a case study in system failure — proof that meaning still exceeds extraction, that ambiguity remains the last human territory.

When the summariser stopped, it wasn’t broken.

It had reached the edge.

So, perhaps it wasn’t baffled—it had just run out of decision tree branches and decided to hide in blankness.

It could summarise the mechanics of MOST of my story; it couldn’t summarise what the story was doing. It’s a theme I’ve dealt with in Artificial and Luka, and most recently in parts of Jellyfish are loading their guns. 

In its way, it feels right. The machine reached the edge, blinked, and stopped.