The Tiered Adjudication System
The appeal right was preserved. Every tier in the system honors it. The upper tiers will hear your objection, consider it, and return a verdict in certified-correct steps you cannot follow — so due process became a filing fee, and accuracy became the argument against accountability.
Overview
The appeal right was preserved. This is important to understand first: nobody removed due process.
When the Tiered Adjudication System was adopted sector by sector between 2155 and 2171, each adoption was justified by the documented failures of the human courts it replaced. The human courts were biased. They were slow. They were expensive. They were producing miscarriages of justice at rates that, once measured, were difficult to defend publicly. The AI tiers were demonstrably better on every metric the courts had been designed to optimize. The transition was sold, accurately, as an improvement.
What the adoption documentation did not include: the practical effect of the transition on the adversarial core of due process. Not the right to appeal — that was preserved. The ability to rebut the reasoning. The capacity to say this inference is wrong and have someone qualified to evaluate whether you were right.
At tier 1, a verdict runs to three or four pages. A human adjudicator's name appears at the bottom. You can read the sentence that condemned you. You can file a brief arguing against that sentence. The system is biased, slow, and expensive, and eleven percent of appeals result in reversal.
At tier 3, a verdict runs to 214 certified pages. Each page is a step in an inference chain certified by Nexus Dynamics as computationally correct. Not just. Not comprehensible. Correct. You can read the document the way you can read weather data: you can look at all the numbers. You cannot argue with them.
The appeals rate at tier 3 is 22%. The reversal rate is 0.3%.
The Five Tiers
Tier 1 — Human Minor: Neighborhood disputes, minor infractions, first-instance employment claims. Human adjudicators, locally appointed. Verdicts in plain language. Appeals to tier 2 or tier 3.
Tier 2 — Human Major: Serious criminal charges, large civil claims, corporate liability. Trained judicial officers under Nexus administrative oversight. Verdicts in legal language — dense, but in principle readable. Appeals to tier 3 or tier 4.
Tier 3 — AI Intermediate: The first AI tier. Handles appeals from tiers 1 and 2, and first-instance adjudication of automated-system torts. Verdicts in certified inference chains: 100-300 pages at tier 3. The Glosser market covers primarily tier 3 verdicts.
Tier 4 — AI Advanced: Appeals from tier 3 and high-stakes civil disputes. Verdicts: 200-800 pages. The inference chains are complex enough that Nexus Dynamics' own documentation acknowledges that "advanced technical expertise" is required to parse individual steps. No public documentation of what "advanced technical expertise" means in this context.
Tier 5 — AI Supreme: Final appeals. Verdicts up to 1,200 pages. In practice, filing a tier 5 appeal requires documentation that the tier 4 verdict contained a procedural error — not a reasoning error, because reasoning cannot be reviewed, but a failure to follow the administrative procedure. Of the 47 tier 5 appeals filed in 2184, all 47 were dismissed for failure to identify a qualifying procedural error.
What "Certified Correct" Means
The certification provided by Nexus Dynamics' Adjudication Infrastructure Division means: given the inputs, the computation proceeded without error to the output. The certification does not address whether the inputs were appropriate, whether the training corpus was representative, whether the outcome is just, or whether a competent human judge would have reached the same conclusion. The certification covers the process. It does not cover the premise.
This distinction — between process and premise, between computational correctness and moral correctness — is the gap the Verdict Gloss market occupies. A Glosser cannot verify their interpretation against the actual chain. A defense advocate cannot challenge the premise without first accessing the chain. The chain is archived in Nexus infrastructure. It has never been disclosed to any defendant's representative.
| Tiers | 5 — tiers 1-2 human-operated, tiers 3-5 AI-operated |
|---|---|
| Reversal Rate Tier 1 | 11% |
| Adoption Period | 2155-2171, sector by sector |
| What Certified Means | Computationally correct — not just, not comprehensible, not rebuttable |
The 0.3%
The 0.3% reversal rate at tier 3 and above is not zero. This is worth acknowledging. Occasionally, an appeal succeeds. When it does, it succeeds on procedural grounds: the AI court failed to follow its own administrative process before issuing the verdict. Not because the verdict was wrong — the certification guarantees that — but because a form was missing, or a notification timing requirement was violated.
No tier 3-5 reversal on the merits has been recorded in the Sprawl as of 2184. No defense advocate has ever successfully argued that the inference chain was incorrect. Nexus considers this the system working as designed.
The Translation Market
The Verdict Gloss is officially recognized under Nexus adjudication standards as a "cognitive accessibility document." This recognition was added to the standards in 2177, two years after the first licensed Gloss firm opened in Sector 1. The recognition requires all licensed Glossers to include in their service agreement a disclosure: This gloss is a rendered interpretation, not a verbatim translation, of the AI adjudication output.
The disclosure is the acknowledgment that the gloss is a fiction. The court system endorsed this acknowledgment. The court system continues to accept Gloss documents as evidence that a defendant received meaningful due process.
The defense advocates who file Habeas Intelligibilis motions — demanding verdicts in readable form before enforcement — have not been able to leverage the disclosure against the court. The arbitrators' consistent ruling: the existence of the Gloss service demonstrates that readable access is available, whether or not the access is accurate. The motion fails. The gloss is cited. Due process is recorded.
The appeals rate at tier 3-5 is 22%. The reversal rate is 0.3%.
How It Works
A case enters the tiered system at the appropriate tier: minor disputes at tier 1, serious charges at tier 2, and so on. Within each tier, adjudication produces a verdict. If dissatisfied, a party may appeal the verdict to the next tier up.
At tier 1 and 2, the appeal path is standard: a human adjudicator in a higher court reviews the reasoning and produces a new, readable verdict. The reasoning of both can be compared. Arguments can be made against it.
At tier 3, the appeal path changes character. The appeal is filed. The fee is paid. The tier 3 court receives the filing. A certified Nexus adjudication shard processes the relevant facts, the applicable precedent, and the disputed reasoning from the tier below. It returns a verdict. The verdict is in certified-correct inference steps. The first step is the input summary. The last step is the conclusion. In between: anywhere from 100 to 1,200 intermediate computation steps, each formally correct, each certified by Nexus's Adjudication Infrastructure Division as proceeding without error from its predecessor.
The verdict is delivered in digital format. It is archived. The archive is stored in Nexus infrastructure. No defense advocate has ever been granted access to review the inference chain against their own case data. The tier 3 verdict is the official record.
If the party files a tier 4 appeal, the tier 4 court receives the new filing and the tier 3 verdict as input. It produces a new verdict, incorporating the tier 3 chain by reference. The tier 4 chain grows longer. The same process applies at tier 5.
Appeal advocates operate under what the adjudication standards call "procedural review": they may argue that the correct administrative process was not followed before the verdict issued, that required documentation was missing, or that the filing was incomplete. They may not argue that the inference chain was wrong. The certification is the answer to that argument.
The certification covers the process, not the premise.
'What does this mean?' The Glosser charges 800 credits and reads them a story.
Connected To
The upper tiers produce verdicts in certified-correct steps no defendant can read; the Verdict Glossers exist because justice requires language and the AI stopped speaking it.
Licensed practitioners producing acknowledged-fiction summaries of AI court verdicts — the service agreement says so.
Nexus certifies every tier 3-5 inference step as correct — unchallengeable in practice, unreadable by design.
Social Impact
The tier 3-5 court system is, by every official measure, a success. Cases resolve faster than in the human tiers. The statistical error rate is lower. Appeals take less human-adjudicator time. Costs to the adjudication system per case are down 60% since full adoption.
The corporations that operate the tiered system report high satisfaction rates among filing parties. The satisfaction measure is whether the party received a verdict — not whether they understood it. On that measure, the system scores nearly 100%.
Outside the official measures, three developments have not appeared in any Nexus administrative report:
The Gloss market. The Verdict Glosser industry emerged two to four years after tier 3-4 adoption in most sectors. It now employs approximately 250 licensed practitioners and an untracked number of independent operators. Its existence is not measured in any adjudication performance metric.
The advocacy attrition. Seventeen advocacy firms that specialized in tier 3-4 appeals have closed since 2175. Their practice areas have not been absorbed by other firms. Defense advocates who attempt to build expertise in tier 3-4 appeals report difficulty sustaining the practice: cases resolve against their clients at rates that make it difficult to argue the practice constitutes representation.
The Boalt Docket's wall. In Sector 11, a free legal clinic has wallpapered its seminar room with enlarged tier 4 verdict pages in 48-point type. The inference steps run floor to ceiling. None of the clinic's nine staff attorneys can parse the single formula circled in red at the center. This is not a secret. It is a condition of practicing in the tiered system that everyone has agreed to stop discussing.
The Verdict Glossers emerged as the market the system made necessary — rendering verdicts the court stopped producing in legible form. The Culpability Market occupies the parallel gap: where Glossers translate verdicts, this market assigns blame to defendants the court declared correct without explaining why. Judge Dreg runs adjudication below the tiers; his court reads defendants rather than processing certified inference chains, which makes it legible but not scalable. Mystery Court uses opacity deliberately — the Court's content is protected; the tiered court's opacity protects the institutional authority. The Severance Pact shares the same structural feature — AI reasoning humans cannot audit — but conceals by design rather than by architecture. Marisol Trejo marks wrong reports wrong and still carries them; the Gloss market's disclaimer structure routes around that marking. The system instantiates Post-Truth Justice as judicial doctrine: certified correctness replaced the right to rebut. Last Human Smarter Than AI is its mechanism: the cognitive ceiling makes inaccessibility a technical condition of legitimate adjudication. Corpo-Nations' accountability architecture extends into enforcement: Nexus certifies correctness, Good Fortune administers contracts, and no single institution answers for what the algorithm decided.