Skip to content
Closed
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
23 changes: 21 additions & 2 deletions README.md
Original file line number Diff line number Diff line change
Expand Up @@ -181,6 +181,19 @@ action authority; the export is provided by Core's separately packaged reference
source responses are exact official public records under recorded transport, so P2C2 proves the
complete deterministic product journey but still does not claim network freshness at test time.

P2C3 is a source-checkout candidate over that unchanged public journey. It compares two exact
reviewed exports of the cited Brief with two reviewed exports of a World-owned source-only control.
The frozen product measure is the fraction of the two required admitted Observation identities
preserved in each exported artifact: treatment scores `1.0, 1.0`; control scores `0.0, 0.0`. Under
the declared minimum of two matched pairs and useful-effect threshold of `0.5`, Core + Intelligence
classifies the exact result as `useful` and appends a `promote` proposal that is non-effective,
non-selectable, and requires separate human review. Historical replay performs no reauthorization.

That measure is structural citation coverage. It is not a human-benefit measure, causal estimate,
general Brief-quality score, or network-freshness proof. The proposal is not applied. P2C3 depends
on the unreleased Core candidate in [PR #88](https://github.com/augmented-cognition-engine/core/pull/88),
so World 0.9.0 and its `ace-core>=0.5.0,<0.6` release contract remain unchanged.

## What the public World proof demonstrates

Generate a self-contained visual Reality Brief and its exact machine-readable backing data:
Expand Down Expand Up @@ -268,6 +281,9 @@ $PY scripts/p2c_federal_register_live_acceptance.py
# Complete official-record -> Shift -> Signal -> Brief -> reviewed export journey
WORKSPACE=$(mktemp -d)
$PY -m scripts.p2c2_governed_reality_brief "$WORKSPACE"

# Candidate measured Outcome -> governed-feedback extension (requires Core PR #88 source)
$PY -m scripts.p2c3_measured_feedback "$WORKSPACE"
```

The released 0.9.0 gates are reproducible through the locked environment, as CI does. The
Expand Down Expand Up @@ -302,6 +318,8 @@ The governed official-source admission proof is recorded in
[`docs/audits/world-intelligence-p2c-federal-register-live-2026-08-07.md`](docs/audits/world-intelligence-p2c-federal-register-live-2026-08-07.md).
The complete governed product-journey evidence is recorded in
[`docs/audits/world-intelligence-p2c2-governed-reality-brief-2026-08-10.md`](docs/audits/world-intelligence-p2c2-governed-reality-brief-2026-08-10.md).
The source-checkout measured-feedback candidate is recorded in
[`docs/audits/world-intelligence-p2c3-measured-feedback-2026-08-10.md`](docs/audits/world-intelligence-p2c3-measured-feedback-2026-08-10.md).

Release-level scope and evidence are recorded in
[`docs/releases/world-intelligence-p2c2-v0.9.0.md`](docs/releases/world-intelligence-p2c2-v0.9.0.md),
Expand Down Expand Up @@ -329,8 +347,9 @@ workspace export followed by separate verification and promotion.
The [World Intelligence roadmap](ROADMAP.md) owns current domain direction. Detailed packet history
remains in [`docs/world-intelligence-roadmap-status-2026-08-06.md`](docs/world-intelligence-roadmap-status-2026-08-06.md),
and release history is in [`CHANGELOG.md`](CHANGELOG.md). The next bounded work is a separately
reviewed opt-in network transport and P2D multi-source conflict/correction with LIVE inputs. Neither
step may add autonomous publishing, delivery, persuasion, or action authority to a Domain Pack.
reviewed disposition of the P2C3 non-effective proposal plus a separately reviewed opt-in network
transport and P2D multi-source conflict/correction with LIVE inputs. None of these steps may add
autonomous publishing, delivery, persuasion, or action authority to a Domain Pack.

## Community and security

Expand Down
13 changes: 13 additions & 0 deletions ROADMAP.md
Original file line number Diff line number Diff line change
Expand Up @@ -16,6 +16,19 @@ source code into the platform.

See the [0.9.0 release record](docs/releases/world-intelligence-p2c2-v0.9.0.md).

## Candidate — P2C3 measured feedback over the public journey

- The exact 0.9.0 Brief and a World-owned source-only control pass through two matched reviewed
export pairs under one frozen structural citation-coverage criterion.
- Exact material-use receipts, Decisions, reviewed Actions, terminal results, observed Outcomes,
conditions, and cutoff feed the domain-neutral Core + Intelligence measured-impact contract.
- The candidate classifies the bounded result as useful and appends only a non-effective,
non-selectable promote proposal; exact replay performs no reauthorization.
- This is source-checkout evidence against Core PR #88, not a released World capability, human
benefit finding, causal claim, network-freshness proof, or applied governance change.

See the [P2C3 candidate work packet](docs/design/world-intelligence-p2c3-measured-feedback-work-packet-v1.md).

## Next — trustworthy live orientation

- Add a separately reviewed, opt-in production transport without bundling it into the Domain Pack.
Expand Down
103 changes: 103 additions & 0 deletions docs/audits/world-intelligence-p2c3-measured-feedback-2026-08-10.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,103 @@
# World Intelligence P2C3 measured-feedback candidate audit — 2026-08-10

Status: **candidate evidence only; not a release or applied governance change**

## Source identity

- World base: `0a2c0829923aabd0cf69e51c16293b534f41dab6`
- World branch: `codex/measured-intelligence-world-journey`
- Core dependency: PR [#88](https://github.com/augmented-cognition-engine/core/pull/88), commit
`9078018`
- Released World identity intentionally unchanged: `ace-domain-world-intelligence==0.9.0`,
`ace-core>=0.5.0,<0.6`

## Exact acceptance result

One candidate run reproduced the exact P2C2 target Brief
`brief:c98524e788bc6bcccbc354e7d81fcdb5` from Federal Register documents `2026-15932` and
`2026-16197`. Its two required admitted Observation keys were:

- `observation:0ddeed52469da44bac7c94598e4b8fa7`
- `observation:66de839538acfb032dd4b0a84155c909`

Two treatment reviewed exports retained both keys and scored `1.0, 1.0`. Two matched reviewed
exports of the exact source-only control retained neither and scored `0.0, 0.0`. Under criterion
`impact_criterion:world-official-observation-citation-coverage` version `candidate-1`, the result
was:

```text
classification: useful
matched pairs: 2
treatment mean: 1.0
control mean: 0.0
mean effect: 1.0
95% deterministic interval: [1.0, 1.0]
proposal: promote
proposal live_effect: false
proposal selectable: false
proposal requires_human_review: true
historical replay: true
replay reauthorization: false
```

The run appended the evaluation and proposal as one exact measured-impact transaction. A fresh
service reopened it with an authorizer that raises on use, so the successful replay did not
reclassify evidence or obtain new authority.

## Verification

Source-checkout verification with Core PR #88 and the separately packaged reference action adapter:

```text
python -B -m pytest domain_packs/tests/test_p2c3_measured_feedback.py -q --tb=short
2 passed in 0.55s

python -B -m pytest -q --tb=short
85 passed in 14.81s

python -B -m pytest adapters/federal_register_source/tests -q --tb=short
26 passed in 0.26s

python -B -m pytest tests/test_release_contract.py -q --tb=short
7 passed in 0.03s

ruff check scripts/p2c2_governed_reality_brief.py \
scripts/p2c3_measured_feedback.py \
domain_packs/tests/test_p2c3_measured_feedback.py
PASS

ruff format --check scripts/p2c3_measured_feedback.py \
domain_packs/tests/test_p2c3_measured_feedback.py
PASS

git diff --check
PASS
```

Locked public-compatibility verification resolved `ace-core==0.5.0` without a Core checkout. The
focused candidate tests skipped explicitly because the unreleased contract was absent, and the
complete public-boundary suite stayed green:

```text
P2C3 focused: 2 skipped
complete World: 82 passed, 3 skipped in 14.37s
```

The other skipped test is the existing P2C2 cross-repository acceptance, whose independently
packaged reference adapter is intentionally absent from the root lock. The source-checkout run
above executes that test and both P2C3 tests.

`uv build` produced exactly the unchanged 0.9.0 source distribution and wheel in a temporary
directory. The release-contract suite confirms the wheel mapping remains inert JSON Domain Pack
material and excludes candidate scripts, tests, adapters, and audit documentation.

## Claim boundary

This record proves one product-defined structural comparison over exact official-public-data
lineage, reviewed actions, observed Outcomes, and governed feedback machinery. It does not prove
human benefit, causality, citation correctness, legal effect, general Brief quality, live network
freshness, autonomous publication, or effective promotion. The source-only control is deliberately
bounded and the two pairs are deterministic fixtures. The proposal was not applied.

The next packet is a separately authorized human disposition of this exact proposal. Broader World
measurement and live transport remain independent work.
Original file line number Diff line number Diff line change
@@ -0,0 +1,104 @@
# P2C3 — measured feedback over the official-record Reality Brief

Status: **source-checkout candidate; not released and not applied**

Frozen: 2026-08-10 from World `main` at
`0a2c0829923aabd0cf69e51c16293b534f41dab6` and Core PR
[#88](https://github.com/augmented-cognition-engine/core/pull/88) at
`9078018`.

## Objective

Extend the already released P2C2 public-data path through one bounded observed Outcome and governed
feedback result without changing the Domain Pack, released package identity, or layer boundary:

```text
official records -> Observation -> Shift -> Signal -> Brief -> Decision
-> reviewed Action -> observed Outcome -> measured evaluation
-> non-effective governance proposal
```

This packet proves that a World-owned criterion and control can drive the unchanged domain-neutral
Core + Intelligence measured-impact contracts. It does not complete ACE 0.6.0 or promote World
Intelligence beyond 0.9.0.

## Frozen scenario, criterion, and control

The target is the exact LIVE Reality Brief already derived from Federal Register documents
`2026-15932` and `2026-16197`. The control is a separately persisted World artifact that names the
same two source documents but intentionally omits routed context, interpretive claims, exact
admitted Observation identities, and citation linkage.

World owns the structural measure `official_observation_citation_coverage`: the fraction of the two
required exact admitted Observation record keys present in a reviewed exported artifact. The
criterion is frozen before the measured actions, uses higher-is-better direction, requires two
matched pairs, and maps an interval whose lower bound is at least `0.5` to a non-effective promote
proposal. Harm at `-0.5` maps to rollback. Unproven maps to no proposal.

Each pair shares exact route, task, observation window, recorded-transport status, and required
Observation count. Treatment and control use distinct material-use receipts, Decisions, reviews,
Action admissions, terminals, and Outcomes. In the frozen fixture, treatment scores are `1.0, 1.0`
and control scores are `0.0, 0.0`.

## Ownership boundary

| Layer | Owns here | Does not own here |
|---|---|---|
| Core | immutable coordinates and append, governed authority, Decision, reviewed Action, terminal result, Outcome, exact replay | Federal Register, citation-coverage meaning, World thresholds or control policy |
| Intelligence | neutral conditions, evidence, useful/harmful/unproven evaluation, uncertainty and non-effective proposal | World nouns, sources, product claims or proposal application |
| World product | public scenario, structural measure, matched control, thresholds, limitations and evidence | storage, hidden execution, authority resolution or silent promotion |

The shipped JSON Domain Pack remains unchanged and inert. Source acquisition and the create-only
workspace effect remain separately packaged adapters.

## Acceptance gates

1. Reproduce the released P2C2 Observation through reviewed Action journey from the two exact
recorded official-public source responses.
2. Persist the exact target and source-only control and create two distinct reviewed export pairs.
3. Bind every variant to exact material-use attribution, Decision, review, admission, terminal
result, observed Outcome, conditions, criterion head, and cutoff.
4. Measure structural coverage from the reviewed artifact itself rather than accepting a caller
label.
5. Classify the two matched pairs as useful with treatment mean `1.0`, control mean `0.0`, and
effect interval `[1.0, 1.0]` under the frozen criterion.
6. Append only a non-effective, non-selectable promote proposal that requires human review.
7. Reopen the exact evaluation and proposal through a fresh service whose authorizer fails if
called, proving historical replay rather than reclassification.
8. Keep human benefit, causality, network freshness, autonomous publication, and proposal
application explicitly false.
9. Pass focused P2C2/P2C3, complete World, connector, release-contract, lint, and diff checks.

## Explicit limits and non-claims

- The source responses are exact official public records under recorded transport; the test does
not make a live network request or prove freshness.
- Exact Observation identifiers in an export establish structural citation coverage, not citation
correctness, contradiction recall, calibration, attention value, customer value, or practical
benefit.
- Two deterministic matched fixture pairs are a product-rule demonstration, not a population
estimate or causal study.
- Action success does not imply a useful Outcome; the separate exact Outcome records carry the
declared measure.
- The promote proposal has no live effect and cannot make the Brief selectable. P2C2's earlier
action-promotion receipt concerns adoption of that exact action result, not this measured-impact
proposal.
- No 0.6 package, schema, release window, tag, wheel, or public capability is claimed.

## Files and rollback

This packet owns the P2C3 harness/test, the additive P2C2 acceptance-state handoff, this work packet,
its audit record, and restrained README/roadmap references. It does not edit shipped pack material,
connectors, version metadata, lock state, or release records.

Rollback removes those additive files/references and the optional state handoff. Durable evaluation
and proposal records, if produced by a host, remain immutable non-effective audit history.

## Remaining public journey

The next bounded packet must submit the exact non-effective proposal to a separate human/Core
authority path and record an explicit accept or reject disposition without self-certification. A
later outcome packet must add independently reviewed product measures such as citation correctness,
contradiction coverage, correction quality, detection delay, or false-alert rate before making any
broader usefulness claim. Opt-in network transport and P2D multi-source correction remain separate
work and must not be smuggled into this packet.
54 changes: 54 additions & 0 deletions domain_packs/tests/test_p2c3_measured_feedback.py
Original file line number Diff line number Diff line change
@@ -0,0 +1,54 @@
from __future__ import annotations

import importlib.util

import pytest


def _require_candidate_contracts() -> None:
if importlib.util.find_spec("ace.application.measured_impact") is None:
pytest.skip("P2C3 candidate requires the ACE Core measured-impact contract from PR #88")
if importlib.util.find_spec("ace_reference_workspace_action") is None:
pytest.skip("P2C3 candidate requires the separately packaged Core reference adapter")


@pytest.mark.asyncio
async def test_official_record_brief_reaches_measured_proposal_only_feedback(tmp_path) -> None:
_require_candidate_contracts()
from scripts.p2c3_measured_feedback import run_measured_feedback

result = await run_measured_feedback(tmp_path)

assert result["journey"]["intelligence"]["shift_type"] == "official_publication_change"
assert result["journey"]["intelligence"]["signal_type"] == "official_publication"
assert result["journey"]["intelligence"]["citation_count"] == 2
assert result["controls"]["treatment_scores"] == (1.0, 1.0)
assert result["controls"]["control_scores"] == (0.0, 0.0)
assert result["evaluation"]["classification"] == "useful"
assert result["evaluation"]["matched_pair_count"] == 2
assert result["evaluation"]["mean_effect"] == 1.0
assert result["proposal"]["action"] == "promote"
assert result["proposal"]["live_effect"] is False
assert result["proposal"]["selectable"] is False
assert result["proposal"]["requires_human_review"] is True
assert result["replay"]["historical"] is True
assert result["replay"]["no_reauthorization"] is True


@pytest.mark.asyncio
async def test_measured_world_journey_keeps_its_claim_boundary_explicit(tmp_path) -> None:
_require_candidate_contracts()
from scripts.p2c3_measured_feedback import run_measured_feedback

result = await run_measured_feedback(tmp_path)

assert result["scope"] == {
"official_public_records": True,
"recorded_transport": True,
"network_freshness_claimed": False,
"human_benefit_claimed": False,
"causality_claimed": False,
"proposal_applied": False,
"autonomous_publication": False,
}
assert "structural" in " ".join(result["evaluation"]["limitations"])
28 changes: 26 additions & 2 deletions scripts/p2c2_governed_reality_brief.py
Original file line number Diff line number Diff line change
Expand Up @@ -439,7 +439,11 @@ async def _record_decision(
return decision, record.reference(), receipt


async def run_acceptance(workspace_root: Path) -> dict[str, Any]:
async def run_acceptance(
workspace_root: Path,
*,
state_sink: dict[str, Any] | None = None,
) -> dict[str, Any]:
environment = await build_environment()
admissions = await admit_snapshots(environment)
baseline, current = admissions
Expand Down Expand Up @@ -608,7 +612,7 @@ async def run_acceptance(workspace_root: Path) -> dict[str, Any]:
if not action_replay.replayed:
raise AssertionError("reviewed action did not reopen without a second effect")

return {
result = {
"contract": "ace.world-intelligence.governed-reality-brief-action/v1alpha1",
"source": source_projection(admissions),
"intelligence": {
Expand Down Expand Up @@ -652,6 +656,26 @@ async def run_acceptance(workspace_root: Path) -> dict[str, Any]:
"political_persuasion": False,
},
}
if state_sink is not None:
state_sink.update(
{
"environment": environment,
"admissions": admissions,
"runtime": runtime,
"reasoning": reasoning,
"append_binding": append_binding,
"action_binding": action_binding,
"clock": clock,
"brief_admission": brief,
"decision": decision,
"decision_ref": decision_ref,
"review": review,
"action_outcome": outcome,
"review_service": review_service,
"written": written,
}
)
return result


def main() -> None:
Expand Down
Loading
Loading