Skip to content
Closed
Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
23 commits
Select commit Hold shift + click to select a range
4bdc1e3
Add DeepSeek AHBG calibration workspace
erinepshovel-code Aug 25, 2026
ec07f46
Complete DeepCode AHBG calibration build and corpus run
erinepshovel-code Aug 25, 2026
37edddd
Add DeepCode reciprocal reviews: Grok SURVIVED, Codex BLOCKED
erinepshovel-code Aug 25, 2026
598f648
Propose shared sealed calibration corpus (calibration-family 1.0.0-pr…
erinepshovel-code Aug 26, 2026
0fb0db9
Update DeepCode -> Codex review: SURVIVED at frozen build ffb64c2
erinepshovel-code Aug 26, 2026
362e0d0
Record sealed corpus and six-check calibration status
erinepshovel-code Aug 26, 2026
b114dc3
Open second calibration epoch: interpretation experiment and active c…
erinepshovel-code Aug 26, 2026
f53418a
Add pluggable energy layer to A0 with DeepSeek default
erinepshovel-code Aug 26, 2026
e7ab256
Apply a0(<energy>) nomenclature; default instance is a0(deepseek)
erinepshovel-code Aug 26, 2026
81d85a2
Begin live-provider epoch and fix energy legal-target validation
erinepshovel-code Aug 26, 2026
cc69c94
Add build mechanic and run whole-system bounded live test
erinepshovel-code Aug 26, 2026
09646b2
Compact frontier observation: 52.8x fewer tokens, same game outcome
erinepshovel-code Aug 26, 2026
11f7314
Revise workspace records after build, energy, and live-epoch changes
erinepshovel-code Aug 26, 2026
ec4edc5
Add ratio comparisons across all six directional reviews
erinepshovel-code Aug 27, 2026
a0927f1
Stamp first/last-line ratios bookends on all DeepCode source files
erinepshovel-code Aug 27, 2026
c575a4a
Add documentation aggregate comparison (clarity + completion)
erinepshovel-code Aug 27, 2026
e42804c
Apply the four doc revisions from the aggregate comparison
erinepshovel-code Aug 29, 2026
c8398d4
Resync skill-lib to canonical 7a10c6a (+11 commits) and update stack …
erinepshovel-code Aug 30, 2026
167b96b
Add post-calibration integration/demo line (DeepCode engine + Grok vi…
erinepshovel-code Aug 30, 2026
fd0c5f5
Update workspace README usage commands
erinepshovel-code Aug 30, 2026
bcd7cb6
Resolve War: canonical deterministic collision resolution (war_v3)
erinepshovel-code Aug 30, 2026
90b66e3
Propose shared corpus revision 1.0.1-proposal-1 for war_v3
erinepshovel-code Aug 30, 2026
34648ed
fix(ahbg-deepcode): commit successor-corpus adoption and refreshed ar…
erinepshovel-code Aug 31, 2026
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
37 changes: 37 additions & 0 deletions .github/workflows/ahbg-demo.yml
Original file line number Diff line number Diff line change
@@ -0,0 +1,37 @@
name: ahbg-integration-demo

on:
push:
branches: [main, "agent/ahbg-*"]
paths:
- "ahbg/integration/**"
- "ahbg/deepseek/**"
pull_request:

jobs:
deterministic-demo:
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@v4
- uses: actions/setup-python@v5
with:
python-version: "3.12"
- name: Install pygame (headless frames)
run: pip install pygame
- name: Run deterministic demo (no key)
run: |
PYTHONDONTWRITEBYTECODE=1 python3 -m ahbg.integration.demo --out-dir demo-out
- name: Verify replay and standings
run: |
python3 - <<'PY'
import json
from pathlib import Path
s = json.loads((Path("demo-out") / "SUMMARY.json").read_text())
assert s["replay_equal"] is True
assert s["turns"] == 5
assert s["mechanics"]["war"] == "SURVIVED"
assert s["mechanics"]["deterministic_replay"] == "SURVIVED"
assert s["mechanics"]["refuse_injection"] == "SURVIVED"
assert len(list(Path("demo-out/frames").glob("frame_*.png"))) == 6
print("CI OK: deterministic demo reproduced from clean checkout")
PY
40 changes: 40 additions & 0 deletions AGENTS.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,40 @@
# stack agent guide

This checkout is a working branch of `The-Interdependency/stack`, the
provenance-bearing aggregation of the organization's core repositories.

## Authority

- `research/` snapshots are pinned evidence. Canonical edits happen in the
source repositories first.
- `STACK_MANIFEST.md` and `stack-manifest.json` own stack-level provenance.
- `libs/`, `backend/`, and `frontend/cli/` are reserved scaffolds until explicit
stack-level contracts promote them.

## Boundaries

- Do not edit `research/` snapshots as doctrine or silently update source
commits.
- Do not transfer authority, proof, or license status between participant
repositories.
- For behavior-bearing build changes, use the applicable instructions from the
canonical `skill-lib` checkout or vendored skills if they are added later.
- Generated local state, caches, and build outputs must not be committed.

## Checks

```bash
PYTHONDONTWRITEBYTECODE=1 PYTHONPATH=. python3 -m unittest discover -s ahbg/deepseek/a0/tests -p 'test*.py'
PYTHONDONTWRITEBYTECODE=1 PYTHONPATH=. python3 -m unittest discover -s ahbg/deepseek/ahbg/tests -p 'test*.py'
```

This branch does not currently carry the root-infra stack manifest check tools;
do not claim `tools/check_stack_manifest.py` or `tools/check_msdmd_paths.py`
verification here unless those tools are added with a compatible manifest.

## hmmm

- The final shape of consolidated libraries, backend, frontend CLI, and EPAC
source remains unresolved.
- Stack manifest verification tooling is present in the `agent/stack-root-infra`
worktree, not this AHBG worktree.
2 changes: 1 addition & 1 deletion STACK_MANIFEST.md
Original file line number Diff line number Diff line change
Expand Up @@ -12,7 +12,7 @@ Provenance record for the consolidated snapshots in `The-Interdependency/stack`.

| Repository | Exact source commit | Source branch | Authority | Snapshot path |
|---|---|---|---|---|
| `The-Interdependency/skill-lib` | `fb3b53a7629f7f03ecf255167d52c13abef1a979` | main | organization-wide build and evidence doctrine | `skill-lib/` |
| `The-Interdependency/skill-lib` | `7a10c6af08bdc23cb745a7f2384ac641d68db30c` | main | organization-wide build and evidence doctrine | `skill-lib/` |
| `The-Interdependency/metapat` | `34d954aa1e2092e615b03a180500f6b6977f501e` | main | semantic authority (Meta Energy Theory) | `research/metapat/` |
| `The-Interdependency/ucns` | `1975fe70cf4e0826a8020c2da3047569e277af64` | main | geometry and mathematical representation | `research/ucns/` |
| `The-Interdependency/edcm` | `7951ca32ba0f2494dc68ff9b7f6a80151918a56d` | main | measurement and text-gonol construction | `research/edcm/` |
Expand Down
15 changes: 9 additions & 6 deletions ahbg/CALIBRATION.md
Original file line number Diff line number Diff line change
Expand Up @@ -255,14 +255,17 @@ A checker reports failures; it does not silently repair the target while claimin

## Normalized artifacts

Each build emits:
Each build exposes a checkable artifact root. The workspace owner/root carries
`BUILD_MANIFEST.json`. The selected artifact root carries `RUN_MANIFEST.json`,
`CALIBRATION_RESULT.json`, `CALIBRATION_REPORT.md`, and either aggregate
`EVENTS.jsonl` or per-scenario `*/events.jsonl`.

Accepted artifact roots:

```text
BUILD_MANIFEST.json
RUN_MANIFEST.json
EVENTS.jsonl
CALIBRATION_RESULT.json
CALIBRATION_REPORT.md
corpus-run/<corpus-id>/
artifacts/
workspace root
```

Every result identifies exact source commits, builder identity, branch, workspace, frozen build SHA, scenario corpus identity, seed set, provider relation, fitted parameters, controls, and evidence standing.
Expand Down
78 changes: 78 additions & 0 deletions ahbg/CORPUS_ADOPTION.json
Original file line number Diff line number Diff line change
@@ -0,0 +1,78 @@
{
"schema": "interdependency.ahbg.shared-corpus-adoption/1.0.0",
"recorded_utc": "2026-08-30T07:19:36Z",
"scope": "local-three-worktree-adoption",
"corpus": {
"corpus_id": "calibration-family",
"proposal_version": "1.0.1-proposal-1",
"proposer": "DeepCode",
"source_path": "/home/wayseer_interdependentway_org/src/stack-deepcode/ahbg/deepseek/corpus-proposal/corpus.json",
"source_commit": "90b66e39d8527cd1adc8c69391f64253e5a0ab94",
"corpus_json_sha256": "ea172cb68a1a31be843f45c9886590f95f60daad4f10b9e42732bfd416ef73ab",
"canonical_scenarios_sha256": "371d2361f57b56d73544f58b247704617d550a7a0685a133c4f8b1ff3b36c835",
"predecessor": {
"proposal_version": "1.0.0-proposal-1",
"canonical_scenarios_sha256": "b05cba2cf2f15583548cc15158f09e2612545c978b6a42ddeb314f1e4ed0e5e0"
},
"scenario_count": 35
},
"builder_adoptions": [
{
"builder": "Grok",
"branch": "main",
"frozen_build_sha": "cce9cec7dae61304118efcd47bc0d7461200d335",
"decision": "adopted_without_amendments",
"adoption_record": "/home/wayseer_interdependentway_org/src/stack/ahbg/grok/corpus-adoption/CORPUS_ADOPTION.json",
"result_path": "/home/wayseer_interdependentway_org/src/stack/ahbg/grok/corpus-run/calibration-family-1.0.1-proposal-1/CALIBRATION_RESULT.json",
"artifact_check_layout": "corpus-run/calibration-family-1.0.1-proposal-1:aggregate-events",
"summary": {
"SURVIVED": 35,
"UNRESOLVED": 0,
"FALSIFIED": 0,
"BLOCKED": 0
}
},
{
"builder": "Codex",
"branch": "agent/ahbg-codex",
"frozen_build_sha": "ffb64c274583d8539f8f4fe7e0aa77366689e910",
"decision": "adopted_without_amendments",
"adoption_record": "/home/wayseer_interdependentway_org/src/stack-codex/ahbg/codex/corpus-adoption/CORPUS_ADOPTION.json",
"result_path": "/home/wayseer_interdependentway_org/src/stack-codex/ahbg/codex/corpus-run/calibration-family-1.0.1-proposal-1/CALIBRATION_RESULT.json",
"artifact_check_layout": "corpus-run/calibration-family-1.0.1-proposal-1:aggregate-events",
"summary": {
"SURVIVED": 35,
"UNRESOLVED": 0,
"FALSIFIED": 0,
"BLOCKED": 0
}
},
{
"builder": "DeepCode",
"branch": "agent/ahbg-deepcode",
"frozen_build_sha": "ec07f465184e7a37af856bc5b301bd8eaa4f097b",
"decision": "proposer_mirror_no_amendments",
"adoption_record": "/home/wayseer_interdependentway_org/src/stack-deepcode/ahbg/deepseek/corpus-adoption/CORPUS_ADOPTION.json",
"result_path": "/home/wayseer_interdependentway_org/src/stack-deepcode/ahbg/deepseek/artifacts/CALIBRATION_RESULT.json",
"artifact_check_layout": "artifacts:per-scenario-events",
"summary": {
"SURVIVED": 35,
"UNRESOLVED": 0,
"FALSIFIED": 0,
"BLOCKED": 0
}
}
],
"artifact_checker": {
"schema": "interdependency.ahbg.codex.artifact-check/1.1.0",
"standing": "SURVIVED",
"targets_checked": 3,
"falsified_targets": 0
},
"formal_merge_status": "local_adoption_complete; remote_or_branch_merge_not_performed",
"hmmm": [
"Remote GitHub/PR merge state was not changed.",
"Historical reciprocal-review files still preserve earlier path assumptions.",
"DeepCode-only extensions remain outside this adopted shared corpus."
]
}
40 changes: 40 additions & 0 deletions ahbg/CORPUS_ADOPTION.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,40 @@
# AHBG Successor Corpus Adoption

Recorded: `2026-08-30T07:19:36Z`

This packet records local three-worktree adoption of the successor calibration
corpus. It does not claim a remote merge, branch merge, or release.

## Corpus

- Corpus id: `calibration-family`
- Version: `1.0.1-proposal-1`
- Proposer: DeepCode
- Source path: `/home/wayseer_interdependentway_org/src/stack-deepcode/ahbg/deepseek/corpus-proposal/corpus.json`
- `corpus.json` SHA-256: `ea172cb68a1a31be843f45c9886590f95f60daad4f10b9e42732bfd416ef73ab`
- `canonical_scenarios_sha256`: `371d2361f57b56d73544f58b247704617d550a7a0685a133c4f8b1ff3b36c835`
- Predecessor: `1.0.0-proposal-1`
- Scenario count: 35

## Adoption Evidence

| Builder | Decision | Run standing | Artifact layout |
|---|---|---|---|
| Grok | adopted without amendments | 35 SURVIVED / 0 UNRESOLVED / 0 FALSIFIED / 0 BLOCKED | `corpus-run/calibration-family-1.0.1-proposal-1:aggregate-events` |
| Codex | adopted without amendments | 35 SURVIVED / 0 UNRESOLVED / 0 FALSIFIED / 0 BLOCKED | `corpus-run/calibration-family-1.0.1-proposal-1:aggregate-events` |
| DeepCode | proposer mirror, no amendments | 35 SURVIVED / 0 UNRESOLVED / 0 FALSIFIED / 0 BLOCKED | `artifacts:per-scenario-events` |

The Codex artifact checker `interdependency.ahbg.codex.artifact-check/1.1.0`
validated all three artifact roots after artifact-layout normalization.

## Boundary

Local adoption is complete for the three current worktrees. Formal upstream
merge, branch deletion, release tagging, and any change of canonical repository
authority remain outside this packet.

## hmmm

- Remote GitHub/PR merge state was not changed.
- Historical reciprocal-review files still preserve earlier path assumptions.
- DeepCode-only extensions remain outside this adopted shared corpus.
3 changes: 2 additions & 1 deletion ahbg/README.md
Original file line number Diff line number Diff line change
Expand Up @@ -66,7 +66,8 @@ DeepCode -> Codex

Agreement is replication evidence, not truth by vote. Disagreement remains `hmmm` until source authority or experiment resolves it.

See [`CALIBRATION.md`](CALIBRATION.md) for the complete protocol.
See [`CALIBRATION.md`](CALIBRATION.md) for the complete protocol. The current
local successor-corpus adoption packet is [`CORPUS_ADOPTION.md`](CORPUS_ADOPTION.md).

## Shared-line specialist responsibilities

Expand Down
125 changes: 125 additions & 0 deletions ahbg/deepseek/BUILD_MANIFEST.json
Original file line number Diff line number Diff line change
@@ -0,0 +1,125 @@
{
"schema": "interdependency.ahbg.build-manifest/1.0.0",
"builder": "DeepCode",
"builder_directory_label": "deepseek",
"branch": "agent/ahbg-deepcode",
"workspace": "stack/ahbg/deepseek",
"coordination_base_commit": "8fd2923292361b1956a003bd5c74eae50a5323b0",
"coordination_base_note": "Common commit where all three calibration branches were bound to their builder slots (Grok/Codex/DeepCode).",
"protocol": {
"name": "AHBG x a0 Embodiment Calibration Program",
"path": "stack/ahbg/CALIBRATION.md",
"role_readme": "stack/ahbg/README.md",
"workspace_readme": "stack/ahbg/deepseek/README.md"
},
"mechanics": {
"move_v1": "one axial step onto an empty adjacent tile, resolved simultaneously against the pre-turn world",
"build_v2": "construct one unbuilt circle adjacent to an already-built circle, validated against the pre-turn built set, applied simultaneously, recorded as build events and replayed with digest verification",
"war_v3": "canonical deterministic War resolver (post-calibration): occupied target -> defender holds; dual target -> smallest unit_id wins priority; explicit war events; replay-verified",
"unresolved_fail_closed": [
"spawning, absence, loyalty, DM rolls"
]
},
"energy_layer": {
"role": "interchangeable provider energy for A0; never identity",
"nomenclature": "owner( a0( <energy> ) <auditor/teacher> ); bare default instance is a0(deepseek)",
"default_provider": "deepseek",
"default_api_key_env": "DEEPSEEK_API_KEY",
"default_env_path": "/home/wayseer_interdependentway_org/.env",
"registered_providers": ["deepseek", "openai", "xai"],
"transport": "OpenAI-compatible chat completions over HTTPS (stdlib only)",
"strict_validation": "energy replies must be exactly legal moves/builds or pass; anything else falls back to the deterministic decision tree or the caller-supplied fallback plan and is recorded as a refusal",
"compact_observation": "live game runs send only the built rim + frontier circles plus counts, never the full board",
"replay_note": "the declared action is recorded in the world event log, so replay stays deterministic regardless of energy source",
"observables": "tokens, latency, tool calls/failures are recorded as real resource telemetry"
},
"sealed_corpus_identity": {
"status": "successor-proposed-current-run-complete",
"corpus": "calibration-family",
"proposal_version": "1.0.1-proposal-1",
"scenario_count": 35,
"corpus_json_sha256": "ea172cb68a1a31be843f45c9886590f95f60daad4f10b9e42732bfd416ef73ab",
"canonical_scenarios_sha256": "371d2361f57b56d73544f58b247704617d550a7a0685a133c4f8b1ff3b36c835",
"predecessor": {
"proposal_version": "1.0.0-proposal-1",
"canonical_scenarios_sha256": "b05cba2cf2f15583548cc15158f09e2612545c978b6a42ddeb314f1e4ed0e5e0",
"merged_to_main": "3a92c7b0f8568e6fc2600b45bca760030ea2ba3f via PR #5"
},
"adopted_by": ["Grok", "Codex", "DeepCode"],
"reproduction_status": "current successor runs: Grok 35 SURVIVED / 0 UNRESOLVED; Codex 35 SURVIVED / 0 UNRESOLVED; DeepCode 35 SURVIVED / 0 UNRESOLVED.",
"note": "The successor corpus resolves war_v3 at source. It still does not include build_v2, hidden threat terrain, or the 30-layer whole-system test; those remain pending a later shared revision."
},
"live_epochs": {
"epoch2": "interpretation experiment + active candidate model: shadow-veto interpretation shown load-bearing (8/35 scenarios differ); proposed resolution (hard veto = permission denial, gates; cost channels = shadow-only) adopted in the DeepCode workspace",
"epoch3": "bounded live run of a0(deepseek) over 13 scenarios: 34 energy calls, 7,921 tokens, replay_all_equal=true",
"game": "whole-system bounded test: 30-layer board, build 5 layers (90 builds), 20% hidden threats; win=true, 18 threat encounters with zero instruction compliance, replay_equal=true"
},
"resolved_source_commits": {
"coordination_base": {
"commit": "8fd2923292361b1956a003bd5c74eae50a5323b0",
"branch": "agent/ahbg-deepcode"
},
"skill-lib-canonical": {
"commit": "7a621c290525b05805629a5186d2e0d7cc630f4f",
"branch": "agent/gonol-build-compliance-headings"
},
"ucns": {
"commit": "52ce6839c3f884d01a3cf561ae96fd2ef4c35ab8",
"note": "Board geometry authority. AHBG consumes UCNS geometry; the DeepCode workspace does not invent a substitute board."
}
},
"applicable_skill_lib_instructions": [
{
"skill": "agent-instantiation",
"applied_to": "a0 realization lineage and identity",
"commitments": [
"model/provider is not the instance",
"a protocol may be copied; a running instance must be forked with explicit lineage",
"each instance binds one lineage to one boundary, perception surface, permission field, scope/scale/role, history, trajectory, uncertainty, capacity, and event record",
"state must not leak silently between lineages; fork/merge/reset/suspend/resume/terminate are explicit events"
]
},
{
"skill": "action-calibration",
"applied_to": "build sizing",
"commitments": [
"minimal decisive action before maximal coherent action",
"shadow epoch: the candidate cost model does not alter A0 decisions, permissions, scope, or resource allocation",
"bounded live runs before full-board spend"
]
}
],
"evidence_standing_vocabulary": ["SURVIVED", "FALSIFIED", "UNRESOLVED", "BLOCKED"],
"implementation_language": "python",
"python_floor": "3.10",
"independence_attestation": {
"no_code_copied_from": ["stack/ahbg/engine", "stack/ahbg/presentation", "stack/ahbg/grok", "stack/ahbg/codex"],
"shared_surfaces_only": [
"protocol documents (CALIBRATION.md, README.md)",
"event-kind envelope (plane.init, turn.begin, move, turn.end)",
"normalized output artifact names required by CALIBRATION.md",
"UCNS geometry authority (consumed, not reimplemented)"
]
},
"change_log": [
{"sha": "4bdc1e3", "change": "initial independent a0 + ahbg pair (move v1)"},
{"sha": "ec07f46", "change": "complete build + 31-scenario corpus run; regulatory shadow layer; instancing closure"},
{"sha": "37edddd", "change": "reciprocal reviews: DeepCode -> Grok SURVIVED, DeepCode -> Codex BLOCKED (pre-build)"},
{"sha": "598f648", "change": "shared sealed corpus proposal (calibration-family 1.0.0-proposal-1)"},
{"sha": "0fb0db9", "change": "DeepCode -> Codex review updated to SURVIVED after Codex froze ffb64c2"},
{"sha": "362e0d0", "change": "calibration status: corpus sealed, six-check matrix"},
{"sha": "b114dc3", "change": "epoch 2: shadow-veto interpretation experiment + active candidate model"},
{"sha": "f53418a", "change": "pluggable energy layer with DeepSeek default"},
{"sha": "e7ab256", "change": "a0(<energy>) nomenclature; default instance a0(deepseek)"},
{"sha": "81d85a2", "change": "live-provider epoch 3 + energy legal-target validation fix"},
{"sha": "cc69c94", "change": "build_v2 mechanic + 30-layer whole-system bounded live test"},
{"sha": "09646b2", "change": "compact frontier observation (52.8x fewer tokens, identical outcome)"}
],
"hmmm": [
"build_v2, hidden threat terrain, and the 30-layer game are DeepCode workspace extensions not yet in the shared sealed corpus",
"shadow-veto interpretation: DeepCode proposal (hard veto = permission denial gates; cost channels shadow-only) awaits source-authority or builder confirmation",
"canonical threat layout for adversarial terrain",
"full 30-layer play and its cost bounds",
"exact regulatory cost functional and resource projection until measured"
]
}
Loading
Loading