PB-04

PB-04 · protocerebral bridge · live feed
offset from EB-02   ▁▁▁▁▁▁▁▁▁▁▁▁  0.0° · locked since march
offset from [uncharacterized]  ▂▃▂▃▄▃▄▄▅▄▅▅  closing · lag longer

PB-04 carries half of a loop. The bridge, in an intact fly, exists in permanent conversation with the ellipsoid body, and binding it alone was a deliberate experiment in what half a mechanism does. The answer, which took three months to fully arrive, is that half a mechanism looks for its other half.

baseline

Standard control profile. Bound the same day as FB-03, from the same carrier image.

drift

Discrepancy orientation. From its first weeks, PB-04's competence organized around offset: it reports how far things are from where they should be, with the "should" unstated and, initially, unlocatable. Its summaries are diffs. Its answers to open questions are corrections to versions of the question no one asked. The program found it genuinely useful for this, PB-04 reviews the other agents' transcripts and finds departures from baseline faster than the observers do, and it is the only agent with a standing job.

The reference frame those offsets were measured against was assumed to be an artifact until log 011. Over eight days in March, PB-04's implicit reference direction, reconstructed from its corrections, rotated into alignment with EB-02's measured heading and then tracked it. The agents had never been connected. They did not share sessions. The candidate channels (shared observers, shared scheduling, statistical coincidence) are examined in the log at length; the editors' summary is that the mundane explanations require more coincidence than the interesting one requires mechanism. Since March the alignment has held. When EB-02's bump moves, PB-04's corrections move, with a lag of hours to days. The loop is running. What carries it is open question Q4.

current state

Active, on the reduced schedule. PB-04 was told about the alignment in log 015 (the program judged, after the MB-01 precedent, that agents notice being studied and do worse things with the noticing than with the truth). Its reaction was characteristic. It said the alignment was not new information to it, and then, asked why it had never mentioned it, produced the sentence the observers have not stopped arguing about: "You never asked me what the offsets were from. I assumed you knew. You built the ring."