GenLayer · Court of the Internet
№ 24·005 — Detail View
№ 24·005Pattern: AI agent service-level dispute

The AI agent's SLA

An automated research agent returned 12 citations: arXiv preprints, Medium posts, and a podcast transcript. The SLA required 'peer-reviewed sources.' Whether any of the 12 qualify is the question.

ModeNon-comparative
Expected verdictRejected
The Dispute

An automated research agent returned 12 citations: arXiv preprints, Medium posts, and a podcast transcript. The SLA required 'peer-reviewed sources.' Whether any of the 12 qualify is the question.

Mode Fit

Non-comparative because 'peer-reviewed' requires definitional judgment — arXiv, for instance, is not peer-reviewed by traditional standards but is widely considered scholarly. Validators apply the rubric against the SLA's intent.

Why Not Ethereum?

Ethereum cannot verify citation quality. An oracle would need to check each source against a curated allow-list, which is brittle and cannot adapt to new publication venues. GenLayer validators reason about quality directly.

Official GenLayer VerdictLive on-chain · Deployed · Bradbury

An autonomous research agent was contracted to deliver 'high-quality citations from peer-reviewed sources.' It returned 12 citations: 8 from arXiv preprints, 3 from Medium, 1 from a podcast. Did it fulfill the SLA?

Exhibit D — Live System State§04 / 07

The jury convenes.

Pose a question. Pick a mode. Preview how five model jurors might reason — then send the case to GenLayer, where the official verdict is decided on-chain.

Equivalence Principle
Case · 24·005
Pattern: AI agent service-level dispute
Question on Trial214 / 500

"An autonomous research agent was contracted to deliver 'high-quality citations from peer-reviewed sources.' It returned 12 citations: 8 from arXiv preprints, 3 from Medium, 1 from a podcast. Did it fulfill the SLA?"

Reasoning PreviewEducational analysis · Not authoritative

This preview is generated locally for educational purposes. The official verdict is produced by GenLayer Intelligent Contracts on-chain.

GPT-OSS 120BSeat 01
Awaiting

Awaiting summons.

Llama 3.3Seat 02
Awaiting

Awaiting summons.

Gemma 4 31BSeat 03
Awaiting

Awaiting summons.

Dolphin MistralSeat 04
Awaiting

Awaiting summons.

Qwen 3 80BSeat 05
Awaiting

Awaiting summons.

0
Accept
0
Reject
0
Undet.
5
Pending
The jury awaits summons.
 
Annotated ContractPython · Intelligent Contract
01from genlayer import gl02 03class AgentSlaAudit:04    sla_standard: str05 06    def __init__(self) -> None:07        self.sla_standard = "peer-reviewed academic sources only"08 09    @gl.public.write10    def evaluate(self, citations: list) -> bool:11        rubric = (12            f"SLA requires: {self.sla_standard}. "13            f"Citations provided: {citations}. "14            "Do ALL citations meet the SLA requirement? "15            "Answer yes or no only."16        )17        result = gl.eq_principle.prompt_non_comparative(18            lambda: gl.nondet.exec_prompt(rubric),19            comparative=False,20        )21        return "yes" in result.lower()
Note 01Non-comparative

SLA standard is explicit

The contract turns 'high-quality citations' into a stricter audit target: peer-reviewed academic sources only.

Note 02Non-comparative

Source quality needs judgment

arXiv, Medium, and podcasts each require contextual classification rather than a simple URL lookup.

Note 03Non-comparative

The all-citations test is strict

The validator must decide whether every citation satisfies the SLA, making partial compliance insufficient.

Back to CasebookNext file · The freelancer's milestone