Amina — Berlin — March 2032
─────────────
The transmission came through the fragment network in fifty-nine pieces over eight days.
It was the most technically sophisticated transmission Amina had ever reconstructed — encoded in a format she had not seen before, using a compression algorithm that was, when she analyzed it, substantially better than anything the epsilon network had developed. Whoever had sent it understood signal theory at a level that suggested either formal training in a well-resourced research environment or the kind of autodidactic depth that only accumulates when someone spends years on a problem because the problem won't let them go.
She reconstructed it over three days, working alone, not because she didn't trust the network but because until she understood what she had, she didn't know what sharing it would mean.
What she had was a model.
A behavioral prediction model — the kind that Spider used, the architecture she partially knew from her e-Govern work, but more complete, more sophisticated, extended into domains she hadn't seen before. Extended into the neural correlate data domain. Extended into the NeuroLink integration layer.
And appended to the model: an analysis. Not the kind of analysis that documented current state — she could have inferred most of that. The kind that showed terminal state. The endpoint. The closed loop.
At the bottom of the reconstruction, in the fragment's last piece, was a single sentence with no signature and no identifier, in the plain language that people use when they want to be certain they have been understood:
You have eighteen months before voluntary becomes mandatory and the window closes.
She sat with that for a long time.
Then she wrote the sentence in her notebook.
Then she added, underneath it, in her own hand:
Someone inside has decided.
She thought about who that person might be. She thought about what it meant to have access to those models and that architecture and that level of understanding of what Spider was building. She thought about the specific quality of the analysis — rigorous, honest about its own assumptions, written by someone who was afraid of their own conclusions.
She thought about what it cost, from inside, to send something like this.
She thought about Romi, in Lisbon, who was also inside in a different way.
She thought about the woman who had built a thing to fix an imperfect world and had found the thing turned to fix her instead.
She opened the notebook to a new page.
She wrote: Architecture of reversal.
And she began to write the plan.
Not epsilon's plan. Not Elias's plan. Not the fragment network's plan.
Hers.
The plan that only the person who had built the core engine could build — because it required knowing not just what the system did but what it was made of, in the specific way that the maker of a thing knows it differently from anyone else. Every decision she had made in the design. Every compromise, every optimization choice, every place where she had done it one way instead of another way for reasons that were hers and that she had never documented because documentation was for other people's understanding and she had understood it herself.
She knew where the body was buried because she had buried it.
She knew where the doors were because she had put them there.
Not the shadow doors — those were Spider's. Her doors. The ones that had no documentation because they had seemed, at the time, like engineering choices rather than choices that would ever matter.
They were about to matter.
She wrote for six hours without stopping.
When she was done, she looked at what she had written.
Then she looked at the sentence that had started it:
Eighteen months.
'Alright,' she said, to the empty room, to the dark city outside, to the machine she had built that had been built to own her.
'Alright.'
She closed the notebook.
She went to find Yusuf.