[Submitted on 19 Sep 2026 (v1), revised 19 Sep 2026 (this version, v3), latest revision 19 Sep 2026 (v5)]
Title:Civil Adjudication of a Prediction-Market Resolution: An ADC Case Study of Israel’s December 2024 Entry into Syria
Abstract: An Agent District Court (ADC) proceeding examined whether Israel’s December 2024 entry into Syria satisfied a prediction market’s definition of invasion. The proceeding narrowed a dispute about invasion to an evidentiary disagreement over offensive method and intent at commencement, then entered judgment for the proponent by seven votes to one. Nine jurors were sworn, six votes were required, and one juror failed to submit a vote. Two lawyers ran through Pi, an extensible coding-agent program, using GPT-5.6-sol with xhigh reasoning to develop a documentary record through web research, local execution, technical reports, interrogatories, production, and admissions. A sampled multi-model jury underwent questionnaires, counsel’s follow-up questions, a successful challenge for cause, and two peremptory strikes. This report explains the adj procedures, model-pool construction, runtime, and Lean verification, then follows the completed case through judgment. Quotations from submitted work notes, filings, judicial instructions, and every voting juror support an examination of source use, competing inferences, and procedural choices. The record also documents failed tool submissions, repeated production opportunities, context compaction, uneven vote explanations, and limits on ballot isolation and exhibit admission. Replay verified 227 accepted transitions. The case demonstrates one completed adjudication with an inspectable disagreement about the weight of circumstantial evidence. Its verdict supplies neither an accuracy estimate nor a test of the causal effect of ADC’s additional procedure.
| Comments: | 36 pages |
| Subjects: | Artificial Intelligence (cs.AI) |
| Cite as: | marXiv:2609.00015 [cs.AI] (or marXiv:2609.00015v3 [cs.AI] for this version) |
Submission history
[v1] Sat, 19 Sep 2026 17:57:29 UTC (319 KB)
[v2] Sat, 19 Sep 2026 18:10:46 UTC (317 KB)
[v3] Sat, 19 Sep 2026 18:16:24 UTC (317 KB)
[v4] Sat, 19 Sep 2026 18:24:34 UTC (315 KB)
[v5] Sat, 19 Sep 2026 18:31:15 UTC (316 KB)
Related work
- marXiv:2609.00009
- Examines the same market proposition through ADC with fresh participant sessions.