← Research archive
Technical note verified Sep 2, 2026

Knowledge Blueprint v004: Deterministic Bootstrap Planning

KESPA Blueprint v004 converted a repaired 729-topic recovered taxonomy plus 37 approved foundational additions into 759 canonical generating topics, 3,982 unique planned card scopes, and 4,807 evidence-source slots. Windows user-environment validation later confirmed the deterministic rebuild matched the expected artifacts byte-for-byte.

knowledge-factory blueprint taxonomy deterministic-planning safety-gates
MARKDOWN

README.md

3,902 bytes SHA-256 9df1925243e7251e…

KESPA-NOTE-001 — Knowledge Blueprint v004: Deterministic Bootstrap Planning

Date: 2026-09-02 Status: Verified — deterministic user-environment validation Project: KESPA AI / NexLabs Studios

What this document records

Before KESPA could build a trusted corpus, it needed a deterministic plan for what knowledge should exist, how many research/evidence scopes should be created, and which topics required stronger review.

Blueprint v004 was that planning artifact.

It did not generate factual knowledge.

Starting point

The recovered taxonomy contained:

  • 729 taxonomy rows
  • 3,810 corrected pre-amendment planned cards
  • 24 audited unsafe classifications
  • 20 domain changes

A bounded foundational-gap amendment then added exactly:

  • 37 approved topics
  • 207 planned card scopes

This produced:

  • 766 total taxonomy rows
  • 4,017 raw planned cards before alias suppression

Canonical result

After suppressing 7 existing aliases, KESPA produced:

  • 759 canonical generating topics
  • 3,982 unique planned card scopes
  • 4,807 planned evidence-source slots

Every planned card remained:

  • production-ready: false
  • selected evidence URLs: 0

No filler was added merely to reach a round corpus-size target.

Safety and review planning

The blueprint marked:

  • 216 high-stakes cards
  • 467 safety-sensitive cards
  • 216 human-review cards
  • 2,975 technical-validation cards
  • 216 cross-check-required cards

This metadata controlled later evidence and verification behavior; it did not itself make any claim trustworthy.

Largest planned domains

  • Web/backend: 912 cards
  • Infrastructure/DevOps: 876
  • Database/data: 378
  • Security: 364
  • AI/ML: 324
  • Business/finance: 308

The approved amendment also introduced narrowly scoped software_engineering and business_operations domains for newly added foundational topics.

Determinism

The historical candidate included deterministic builders and validators.

The archived local validation recorded:

  • all pipeline scripts compiled successfully;
  • the exact 729-row baseline hashes were checked;
  • the approved amendment count was exactly 37;
  • final blueprint IDs were unique;
  • major machine-readable artifacts reproduced byte-for-byte;
  • explicit LF newlines were later enforced across operating systems so deterministic SHA-256 comparisons would survive Windows newline translation.

Why this mattered

The purpose was not to maximize corpus size.

The purpose was to define an auditable research plan before spending provider tokens or generating persistent knowledge. Known foundational gaps were corrected before evidence acquisition so later retrieval/answer benchmarks would not be biased by intentionally missing basic concepts.

Important boundary

This note documents a planning-stage artifact.

At this point:

  • no factual production cards had been generated;
  • no evidence had been fetched;
  • no provider/model calls were made;
  • production Chroma was unchanged;
  • Brain was unchanged;
  • the database was unchanged.

The original public record was conservatively marked published because the first source archive did not yet contain the separate user-environment validation result.

A later KESPA changelog/Jira record supplied that missing result:

  • Windows Python: 3.11.9
  • deterministic v004 rebuild: byte-for-byte match
  • blueprint counts: unchanged
  • production state changes: none
  • final validator result: [PASS] Windows validation completed successfully

Because the missing user-environment validation was later supplied, this public record is now correctly marked verified.

That verified status applies to the deterministic blueprint artifact and its reproduction, not to the factual truth of future cards that the blueprint planned.

TEXT

SHA256SUMS.txt

393 bytes SHA-256 a473520c5e92fce5…
9df1925243e7251ef60f40f5c5d89b3b06e2bd23acdd466baa2099ccb87173da  README.md
2160c5f68f008e2f5ad615715f69f189e36f02037913834637d2cefecd185d55  note.json
d8119e53749121ee44c1e907736dc96159a32e2bff72a8c449c293a91dd4e74a  metrics.csv
ad853def1764aaa46d52285c3674c360b1ea34eda085801dd4ad0c6b102987ec  methodology.md
1e558918a80f81379a6c7ba53607973a86ef33eb444c3a2d0d9df4f131adcc49  provenance.json
MARKDOWN

methodology.md

1,377 bytes SHA-256 ad853def1764aaa4…

Methodology

Pipeline sequence

Blueprint v004 used a deterministic planning pipeline:

1. verify the exact recovered 729-row manifest and knowledge-plan inputs;

2. reapply only the previously audited unsafe-classification corrections;

3. add exactly 37 approved foundational topics;

4. confirm that existing corrected rows were not altered by the amendment;

5. preserve all taxonomy rows while suppressing 7 high-confidence aliases from card generation;

6. generate canonical card-level research/evidence scopes;

7. generate pending evidence-source slots;

8. validate IDs, counts, amendment membership, policy inheritance, alias suppression, and production-readiness state.

Design rule

The blueprint separated planning from truth.

A planned card described what should eventually be researched and verified. It was never treated as a factual knowledge record merely because a deterministic planner created it.

Deterministic validation

The archived package included macOS and Windows validation paths and byte-for-byte comparison of major generated machine-readable artifacts.

A later validator revision changed generated text/JSONL writing to explicit LF line endings across supported operating systems so that newline translation would not create false SHA-256 mismatches.

The blueprint content and counts did not change as part of that newline fix.

CSV

metrics.csv

1,184 bytes SHA-256 d8119e53749121ee…
metric,value,meaning
recovered_taxonomy_rows,729,baseline taxonomy rows
audited_unsafe_classifications,24,previously audited unsafe classifications
domain_changes,20,classification repair domain changes
corrected_pre_amendment_cards,3810,"planned cards after repair, before amendment"
approved_foundational_additions,37,bounded amendment topics
added_planned_cards,207,planned card scopes added by amendment
amended_taxonomy_rows,766,taxonomy rows after amendment
amended_raw_planned_cards,4017,planned cards before alias suppression
aliases_suppressed,7,existing high-confidence aliases
canonical_generating_topics,759,topics generating final card scopes
final_unique_card_scopes,3982,canonical planned cards
evidence_queue_slots,4807,planned evidence-source slots
high_stakes_cards,216,cards requiring high-stakes policy
safety_sensitive_cards,467,cards marked safety-sensitive
human_review_cards,216,cards requiring human review
technical_validation_cards,2975,cards requiring technical validation
cross_check_required_cards,216,cards requiring independent cross-check
production_ready_cards,0,planning-stage result
selected_evidence_urls,0,no evidence selected at blueprint stage
JSON

note.json

2,988 bytes SHA-256 2160c5f68f008e2f…
{
    "schema": "kespa.public_technical_note.v1",
    "id": "KESPA-NOTE-001",
    "title": "Knowledge Blueprint v004: Deterministic Bootstrap Planning",
    "date": "2026-09-02",
    "status": "verified",
    "purpose": "Document the planning-stage design and measured output of the KESPA v004 knowledge blueprint before evidence acquisition or production knowledge generation.",
    "baseline": {
        "recovered_taxonomy_rows": 729,
        "audited_unsafe_classifications": 24,
        "domain_changes": 20,
        "corrected_pre_amendment_cards": 3810
    },
    "foundational_gap_amendment": {
        "approved_topics_added": 37,
        "planned_cards_added": 207,
        "amended_taxonomy_rows": 766,
        "amended_raw_planned_cards": 4017,
        "new_domains_for_amendment_topics": [
            "software_engineering",
            "business_operations"
        ]
    },
    "canonical_blueprint": {
        "canonical_generating_topics": 759,
        "aliases_suppressed": 7,
        "final_unique_card_scopes": 3982,
        "evidence_queue_slots": 4807,
        "production_ready_cards": 0,
        "selected_evidence_urls": 0
    },
    "policy_counts": {
        "high_stakes_cards": 216,
        "safety_sensitive_cards": 467,
        "human_review_cards": 216,
        "technical_validation_cards": 2975,
        "cross_check_required_cards": 216
    },
    "target_card_distribution": {
        "4_cards_per_topic": 245,
        "5_cards_per_topic": 89,
        "6_cards_per_topic": 432
    },
    "largest_domain_card_counts": {
        "web_backend": 912,
        "infra_devops": 876,
        "database_data": 378,
        "security": 364,
        "ai_ml": 324,
        "business_finance": 308
    },
    "safety_and_data_impact": {
        "network_or_provider_calls": 0,
        "production_chroma_changed": false,
        "brain_changed": false,
        "database_changed": false,
        "factual_cards_generated": false
    },
    "validation_status": {
        "assistant_side_local_artifact_validation": "passed",
        "deterministic_byte_for_byte_rebuild": "passed",
        "windows_user_environment_validation": "passed",
        "windows_python_version": "3.11.9",
        "windows_validation_result": "[PASS] Windows validation completed successfully",
        "content_counts_changed_by_validation": false,
        "production_state_changed_by_validation": false,
        "publication_status_reason": "Upgraded from published to verified after the later user-environment record confirmed successful Windows validation and byte-for-byte deterministic rebuild behavior."
    },
    "limitations": [
        "The blueprint is a planning artifact, not a factual knowledge corpus.",
        "No source URLs or evidence had been selected at this stage.",
        "No card was production-ready.",
        "The 37-topic amendment was intentionally bounded and was not intended to be an exhaustive enterprise IT ontology.",
        "Existing corrected taxonomy topics were not broadly reclassified into the two new amendment domains.",
        "Verified status here refers to deterministic artifact validation in the user environment, not factual verification of the planned future knowledge cards."
    ]
}
JSON

provenance.json

2,127 bytes SHA-256 1e558918a80f8137…
{
    "schema": "kespa.public_provenance.v1",
    "research_id": "KESPA-NOTE-001",
    "historical_source_archive": "forge_ai_research.zip",
    "source_artifacts": [
        {
            "path": "FORGE_blueprint_v004_candidate/README_VALIDATION.txt",
            "sha256": "32b9bd38708d03d3d208a87589e08060f86cccb97c1f9248276aeb21bb8a1dc7",
            "bytes": 4714
        },
        {
            "path": "FORGE_blueprint_v004_candidate/LOCAL_ARTIFACT_VALIDATION.txt",
            "sha256": "db740ef1a3ac5f469ec47c0a2faaeabe8e7abe102e26f3ab4b08de11fda4f7f7",
            "bytes": 1195
        },
        {
            "path": "FORGE_blueprint_v004_candidate/AUDIT_CHANGELOG.txt",
            "sha256": "8b3eb5d20f18dc468f0f3618e27b31c656579fc093aed6860c5f658166a43c92",
            "bytes": 4828
        },
        {
            "path": "FORGE_blueprint_v004_candidate/amendment/foundational_gap_amendment_summary_v004.txt",
            "sha256": "06fa9a914be30cf73c09640da9c20a8e9ee8b27438300d999107d82215780215",
            "bytes": 1022
        },
        {
            "path": "FORGE_blueprint_v004_candidate/input_repair/blueprint_input_repair_summary_v002.txt",
            "sha256": "c71d87479fb2bec8393a29fa3b0e92a981c040b15ceb7a726200544d91de7dc7",
            "bytes": 1442
        },
        {
            "path": "FORGE_blueprint_v004_candidate/full_blueprint/full_blueprint_summary_v004.txt",
            "sha256": "920ea5b210c822fc1723d379e207403c922a8497a4566a99892364089b2080c4",
            "bytes": 4412
        }
    ],
    "publication_note": "This public note publishes derived counts, methodology, limitations, source hashes, and the later user-environment validation outcome. The full historical blueprint JSONL/CSV corpus, implementation scripts, private repository metadata, and internal candidate package remain intentionally unpublished.",
    "status_update": {
        "previous_public_status": "published",
        "new_public_status": "verified",
        "reason": "Later user-environment validation record supplied after the original public package was created.",
        "windows_python_version": "3.11.9",
        "deterministic_rebuild": "byte-for-byte match",
        "validator_result": "[PASS] Windows validation completed successfully",
        "production_changes": "none"
    }
}