Experiment verified Sep 15, 2026

Autonomous Knowledge Gap Acquisition

KESPA detected a live knowledge gap during normal use, researched the missing concept from multiple independent public sources, verified four claims, promoted the result to trusted knowledge, indexed it for retrieval, and recorded provenance through NexLedger.

adaptive-learning knowledge-gap evidence-verification trusted-knowledge nexledger

MARKDOWN

README.md

2,086 bytes SHA-256 65a20152647f364d…

KESPA-EXP-001 — Autonomous Knowledge Gap Acquisition

Date: 2026-09-15 Status: Verified Project: KESPA AI / NexLabs Studios

Research question

Can KESPA identify a missing stable concept during normal use, gather independent public evidence, verify candidate claims, and promote the result into trusted retrieval knowledge without retraining the base model?

Live test

The user asked:

What is a Cuckoo filter?

At the time of the request KESPA recorded:

Trust score: 0.480
RAG cards: 0
Top RAG score: 0.000

The natural research gap detector classified the interaction as a genuine public-general-knowledge gap and created a research request.

Result

The isolated research lifecycle completed successfully:

3 independent source organizations
3 fetched evidence records
4 candidate claims
4 verified claims
Trusted release: forge-release-07990ffc7fc9d09f9be2dd3b
Candidate: #82
Verification: PASS
Trusted Knowledge: #74
Indexed for retrieval: yes
NexLedger asset: FORGEAI#TK123V1

What this demonstrates

This run demonstrates that KESPA can, for at least one stable public-knowledge case:

1. detect weak or absent retrieval coverage during normal use;

2. create a research request automatically;

3. discover and fetch public evidence;

4. require multiple independent organizations;

5. generate candidate factual claims;

6. verify those claims against supplied evidence;

7. produce a trusted release;

8. promote the release to production trusted knowledge;

9. index that knowledge for future retrieval; and

10. establish public provenance through NexLedger.

The base local model weights were not changed.

What this does not prove

This experiment does not establish broad autonomous-learning reliability, general superiority over larger models, or a quantified improvement in answer quality. A separate controlled before/after retrieval study is still required.

Public provenance

NexLedger:

https://nexledger.nexlabs.studio/explorer/asset?asset=FORGEAI%23TK123V1

TEXT

SHA256SUMS.txt

240 bytes SHA-256 c2db28ec16a32ce1…

e6ee930c6f64a923d3ec3735ece99620e7da7d6b6bcf29d46bd6f1e356854370  experiment.json
65a20152647f364d7123ed6de7e8b249dbb11c707d5993589666f1ea756a8db1  README.md
2cf120e83187a68ff5fb3c1ada0a808f773b8bf909abc917b4bca54feb7d4a88  provenance.json

JSON

experiment.json

2,094 bytes SHA-256 e6ee930c6f64a923…

{
    "schema": "kespa.public_experiment.v1",
    "id": "KESPA-EXP-001",
    "title": "Autonomous Knowledge Gap Acquisition",
    "date": "2026-09-15",
    "status": "verified",
    "research_question": "Can KESPA identify a missing stable concept during normal use, gather independent public evidence, verify candidate claims, and promote the result into trusted retrieval knowledge without retraining the base model?",
    "test_prompt": "What is a Cuckoo filter?",
    "initial_state": {
        "trust_score": 0.48,
        "rag_cards": 0,
        "top_rag_score": 0,
        "gap_detected": true
    },
    "research_result": {
        "independent_source_organizations": 3,
        "fetched_evidence_records": 3,
        "candidate_claims": 4,
        "verified_claims": 4,
        "trusted_release_id": "forge-release-07990ffc7fc9d09f9be2dd3b",
        "trusted_release_sha256": "07990ffc7fc9d09f9be2dd3b5594285e5ee0789f7430e24fb1a812cbbf996ed3"
    },
    "production_result": {
        "candidate_id": 82,
        "candidate_status": "promoted",
        "verification_status": "pass",
        "trusted_knowledge_id": 74,
        "indexed": true
    },
    "provenance": {
        "nexledger_asset": "FORGEAI#TK123V1",
        "nexledger_url": "https://nexledger.nexlabs.studio/explorer/asset?asset=FORGEAI%23TK123V1"
    },
    "runtime": {
        "primary_local_hardware": "NVIDIA RTX 3070 8GB",
        "research_model_provider": "Groq",
        "research_model": "openai/gpt-oss-120b",
        "groq_requests": 2,
        "groq_tokens_total_across_calls": 11034
    },
    "scope": {
        "classification": "public_general_knowledge",
        "stable_concept": true,
        "private_user_context_used": false
    },
    "limitations": [
        "This is a single live end-to-end production proof, not evidence of broad autonomous-learning reliability.",
        "The experiment demonstrates the mechanism and lifecycle for one stable public-knowledge topic.",
        "Search/provider failures occurred during development of the worker before the final successful run.",
        "A separate controlled before/after retrieval-quality study is still required to quantify the improvement caused by the newly indexed knowledge."
    ]
}

JSON

provenance.json

429 bytes SHA-256 2cf120e83187a68f…

{
    "schema": "kespa.public_provenance.v1",
    "experiment_id": "KESPA-EXP-001",
    "trusted_release_id": "forge-release-07990ffc7fc9d09f9be2dd3b",
    "trusted_release_sha256": "07990ffc7fc9d09f9be2dd3b5594285e5ee0789f7430e24fb1a812cbbf996ed3",
    "candidate_id": 82,
    "trusted_knowledge_id": 74,
    "nexledger_asset": "FORGEAI#TK123V1",
    "nexledger_url": "https://nexledger.nexlabs.studio/explorer/asset?asset=FORGEAI%23TK123V1"
}