◎ OS PUB Apache 2.0 ← All specifications

P138 — AIEP — Autonomous Research Loop Protocol

Publication Date: 2026-04-12 Status: Open Source Prior Art Disclosure Licence: Apache License 2.0 Author/Organisation: Phatfella Ltd Schema: AIEP_OS_SPEC_TEMPLATE v1.0.1 — https://aiep.dev/schemas/aiep-os-spec-template/v1.0.1

Field of the Invention

[0001] The disclosure relates to autonomous evidence retrieval and synthesis systems within computing environments.

[0002] More particularly, the disclosure concerns an autonomous research loop protocol for use within an Architected Instruction and Evidence Protocol (AIEP) system, enabling an AI agent to iteratively retrieve evidence, evaluate admissibility, identify gaps, and refine retrieval queries in a deterministic governed loop with explicit termination conditions.

Framework Context

[0003] This invention operates within an Architected Instruction and Evidence Protocol (AIEP) environment as defined in United Kingdom patent application number GB2519711.2, filed 20 November 2025, the entire contents of which are incorporated herein by reference.

[0004] The present invention extends iterative evidence refinement mechanisms defined in P40, evidence gap detection mechanisms defined in P16, and AI agent evidence query optimisation mechanisms defined in P15 while remaining independently implementable as described herein.

Background

[0005] Answering complex queries requires multiple sequential evidence retrieval cycles where each cycle informs the next. Without explicit governance, autonomous research loops may execute indefinitely, override admissibility constraints mid-loop, or terminate on heuristic confidence thresholds that are not reproducible.

[0006] There exists a need for a deterministic autonomous research loop with governance-bound cycle limits, explicit admissibility evaluation at each cycle, cryptographically bound cycle records, and fail-closed termination when cycle limits are reached without admissible evidence.

Summary of the Disclosure

[0007] A computer-implemented autonomous research loop protocol comprises: receiving a ResearchQuery specifying a subject, an admissibility threshold, and a maximum cycle count; executing a ResearchCycle comprising evidence retrieval, admissibility evaluation, and gap identification; on each cycle, computing a CycleStateHash as H(cycle_index || retrieved_evidence_hashes || admissibility_result || gap_descriptor); appending a ResearchCycleRecord binding cycle index, cycle state hash, retrieved evidence hashes, and admissibility outcome; if the admissibility threshold is satisfied, terminating the loop and returning the admissible evidence set with its cycle records; if the maximum cycle count is reached without satisfying the admissibility threshold, terminating the loop and returning a non-admissible outcome with all cycle records; and prohibiting loop continuation beyond the governance-bound cycle limit.

[0008] The technical effect is modification of computing system behaviour by enforcing deterministic bounded evidence retrieval loops with cryptographically auditable cycle records, preventing unbounded autonomous execution.

Claims

[0009] A computer-implemented method for autonomous research loop execution comprising: receiving a ResearchQuery with admissibility threshold and maximum cycle count; executing evidence retrieval and admissibility evaluation cycles; computing a CycleStateHash per cycle; appending ResearchCycleRecords throughout the loop; terminating on admissibility satisfaction or cycle limit; and returning results with audit records.

[0010] The method of claim 1 wherein termination on cycle limit exhaustion without admissibility satisfaction produces a non-admissible outcome without heuristic override.

[0011] The method of claim 1 wherein each CycleStateHash binds cycle index, retrieved evidence hashes, and admissibility result into a single verifiable commitment.

[0012] A system for autonomous research loop execution comprising one or more processors and a non-transitory computer-readable medium storing instructions to execute the method of any preceding claim.