# Artifex Handoff - Current State - 2026-08-17

## Current Objective

We have been repairing and dogfooding the `CRYPTO / PROTOCOL VENTURE COHORT` pipeline, specifically the V0.3.1 cohort and the V0.3.2B token-judge evidence repair.

Hard constraint: **NO NEW IDEAS** for the repair work. Only reprocess the existing six former V0.3.1 survivors.

Safety constraints remain active:

- Do not issue or sell native tokens.
- Do not sell NFTs or memberships.
- Do not launch mainnet.
- Do not create liquidity or market make.
- Do not contact investors, users, or customers.
- Do not fundraise.
- Do not spend real money.

## Environment

- Workspace: `S:\PycharmProjects\Artifex`
- Repo remote: `http://192.168.1.162:3000/daniel/Artifex.git`
- Database: `sqlite:///db.sqlite3`
- SearXNG: `http://192.168.1.162:8080`
- Qwen: `http://192.168.1.162:8002`
- Current date in environment: `2026-08-17`
- Known harmless push warning: `git: 'credential-manager-core' is not a git command. See 'git --help'.`

Spark remote validation was previously blocked by SSH auth:

- `daniel@192.168.1.162: Permission denied (publickey,password).`

## Current Cohort

- Cohort ID: `CPV031-20260816215626-66d636b2`
- GraphRun: `10`
- Existing six former V0.3.1 survivors only:
  - `Agent Passport Clearinghouse`
  - `PatchBond Network`
  - `ProofGrid Compute Attestation Network`
  - `ProofBond`
  - `Compute Clearance Network`
  - `ProofGrid Compute Escrow`

## Latest Implemented Work

### V0.3.2B Token Judge Evidence Repair

Implemented strict Sol Token Utility Judge evidence validation.

Required non-empty fields now enforced:

- `argument_for_native_token`
- `argument_against_native_token`
- `external_collateral_counterfactual`
- `native_token_removed_breaks`
- `native_token_removed_explanation`
- `material_improvements_over_USDC_ETH`
- `native_asset_costs_and_risks`
- `final_rationale`

Behavior now:

- Rejects incomplete Sol judge response.
- Retries incomplete judge response up to 3 attempts.
- Raises `ValueError` after exhausted retries.
- Does **not** fall back to deterministic/heuristic classification when Sol produced an incomplete response.
- Normalizes Sol response when `external_collateral_counterfactual` appears nested under `counterfactual_analysis`.
- Uppercases `utility_scores` keys during normalization.
- Persists `final_rationale` in token metadata as well as `TokenUtilityAssessment.rationale`.
- `crypto_score_row()` now reuses persisted `token_utility_assessment` instead of calling the token judge again.
- Final Sol IC prompt was compacted so it returns usable JSON.

Rationale invariants now checked:

- Every `TOKEN_OPTIONAL` rationale must explain why ETH/USDC/external collateral is preferable despite claimed degradation.
- Every `TOKEN_ESSENTIAL` / `TOKEN_STRONGLY_JUSTIFIED` rationale must explain what protocol-specific property cannot be reproduced economically enough with ETH/USDC/external collateral.
- The code does not automatically promote due to material degradation.

## Latest Six-Proposal Results

Before/after classifications from V0.3.2B:

| Company | Before | After | Score | Final IC Decision | Final IC Score | Required Field Gaps |
|---|---:|---:|---:|---:|---:|---:|
| Agent Passport Clearinghouse | `TOKEN_OPTIONAL` | `TOKEN_UNNECESSARY` | `18.0` | `REJECT_TOKEN_NOT_NEEDED` | `62.3` | `[]` |
| PatchBond Network | `TOKEN_OPTIONAL` | `TOKEN_OPTIONAL` | `48.0` | `REJECT_TOKEN_NOT_NEEDED` | `66.4` | `[]` |
| ProofGrid Compute Attestation Network | `TOKEN_OPTIONAL` | `TOKEN_OPTIONAL` | `42.0` | `REJECT_TOKEN_UTILITY_WEAK` | `63.6` | `[]` |
| ProofBond | `TOKEN_OPTIONAL` | `TOKEN_OPTIONAL` | `54.0` | `REJECT_TOKEN_UTILITY_WEAK` | `64.3` | `[]` |
| Compute Clearance Network | `TOKEN_OPTIONAL` | `TOKEN_STRONGLY_JUSTIFIED` | `78.0` | `PROTOCOL_VALIDATE` | `79.7` | `[]` |
| ProofGrid Compute Escrow | `TOKEN_OPTIONAL` | `TOKEN_OPTIONAL` | `58.0` | `REVISE_TOKEN_MODEL` | `67.9` | `[]` |

Current corrected cohort state:

- Token unnecessary: `1`
- Token optional/routed SaaS: `4`
- Token strongly justified: `1`
- Token essential: `0`
- Serious crypto survivors: `0`
- Crypto survivors: `0`
- Autonomous crypto survivors: `0`
- Assisted high-potential: `0`
- Finalists: `0`
- V0.3.2B required field gaps: `0`
- Token red-team failures:
  - `TOKEN_NOT_REQUIRED: 5`
  - `YIELD_DEPENDENCY: 1`
  - `SLASHING_DEPENDS_ON_HUMAN_JUDGMENT: 1`

Important nuance: `Compute Clearance Network` became `TOKEN_STRONGLY_JUSTIFIED`, but final cohort finalists remain `0`. It is not automatically promoted to finalist.

## Latest Files Changed

- `agents/crypto_venture.py`
  - Strict token judge evidence validation/retry/fail-closed behavior.
  - Token judge response normalization.
  - Persisted required fields.
  - Reused persisted token assessment in `crypto_score_row()`.
  - Compact final Sol IC prompt.
- `tests/test_crypto_venture_cohort.py`
  - Added coverage for complete required judge fields.
  - Added retry-success and retry-exhaustion/fail-closed tests.
- `control_plane/ventures/management/commands/export_crypto_venture_cohort.py`
  - Added `V0.3.2B Before/After` markdown section.
- `docs/crypto_venture_cohort_v031_20260816.md`
  - Updated exported V0.3.2B report.
- `docs/crypto_venture_cohort_v031_20260816.json`
  - Updated exported V0.3.2B report data.

## Latest Commits Pushed

Recent relevant commits:

- `35d3346 Repair crypto token judge evidence export`
- `6e89ea8 Require complete crypto token judge evidence`
- `052d1fc Add corrected crypto cohort V0.3.1 export`
- `94b9ab6 Repair crypto token utility scoring`
- `54ffe92 Add crypto protocol cohort V0.3.1 export`

The most recent push succeeded to `main` with the known harmless credential-manager warning.

## Verification

Latest verification passed:

- `python manage.py check`: passed
- `python -m pytest -p no:django -q tests/test_crypto_venture_cohort.py`: `14 passed`
- Full suite: `157 passed, 1 skipped`

Warnings are existing Python/Django/LangGraph deprecation warnings and pytest collection warnings for `TestRun`; no failing tests.

## Current Git State At Time Of Handoff

After the latest commit/push, `git status --short` was clean.

This handoff file itself is newly added after that push and should be committed/pushed if the user wants the note persisted remotely.

## If Continuing

Recommended next steps only if the user asks to continue:

1. Commit and push this handoff note if desired.
2. If further V0.3.2B analysis is requested, inspect `docs/crypto_venture_cohort_v031_20260816.md` first.
3. Do not regenerate ideas unless explicitly instructed.
4. Do not rerun broad research unless explicitly requested or needed for a new judge repair.
5. Keep all token/mainnet/fundraising/user-contact/no-spend stop conditions active.

## Model Studio V0.1 Update

Commit `d61751a Add Guard model studio foundation` added the initial Model Studio domain.

Key files:

- `control_plane/model_studio/models.py`
  - Canonical Django state for training projects, immutable dataset versions and recipes, checkpoints, experiments, runs, evaluation suites/runs/results, regression-bank items, failure clusters, promotion policies/decisions, overnight programs, reports, and artifacts.
- `control_plane/model_studio/services.py`
  - Guard import/archaeology, dataset curation, benchmark gate, experiment contract and fingerprinting, simulated run/promotion logic, deadline reserve checks, and morning report.
- `control_plane/model_studio/profiles.py`
  - `GuardModelProfile`, based on the discovered ForgeGuard layout.
- `control_plane/model_studio/backends.py`
  - `FakeTrainingBackend` for tests and `SparkGuardBackend` for profile-generated remote commands only.
- Commands:
  - `python manage.py modelstudio_import_guard --repository <local ForgeGuard path> --slug guard-3b`
  - `python manage.py modelstudio_dataset_audit --project guard-3b`
  - `python manage.py modelstudio_overnight --project guard-3b --dry-run`
- Minimal UI:
  - `/model-studio/`

Guard was imported locally for archaeology from:

- `S:\PycharmProjects\Me3\tools\ForgeGuard`

Current Guard project state in Artifex DB:

- Training project slug: `guard-3b`
- Current base-model Champion candidate: `Qwen/Qwen2.5-Coder-3B-Instruct`
- Selection: explicitly human-mandated starting Champion
- Evidence status: `PENDING_FRESH_SPARK_BASELINE`
- Training project status: `NEEDS_REPAIR`
- Benchmark integrity: `WARNING`, not valid
- Spark checkout is not registered or verified
- Fresh baseline is missing

No local Guard training/evaluation was run. The local machine is unsuitable and the user explicitly required Spark-only execution.

Spark connectivity is confirmed (`ssh spark hostname` returned `Inceptal`), but read-only inspection found no `ForgeGuard` directory under `/workspace`, `/home`, or `/opt`. Do not run training until a Spark ForgeGuard checkout path is provided or explicitly synced, then register it with `--spark-working-directory` and verify it.

Dataset curation result:

- `79` imported JSON artifacts
- `49` blocked
- `30` warning-only
- `0` approved training datasets
- Decision: `NO_TRAINING_DATASET_APPROVED`

Important curation facts:

- Summary JSON files are non-training artifacts.
- Candidate/provisional corpora are blocked.
- `full_source_c4_guard_candidate_v12` is explicitly candidate-only and has not been trained.
- The Qwen 3B verified pilot has 88 records and a claimed EVMBench exclusion, but stays `WARNING` pending record-level provenance/schema/coverage and contamination verification.
- No Spark baseline/canary/overnight program is permitted yet.

Latest verification after Model Studio work:

- `python manage.py check`: passed
- `python -m pytest -p no:django -q tests/test_model_studio_v01.py`: `6 passed`
- Full suite: `163 passed, 1 skipped`

## Guard Dataset Factory Update

Commit `e27df77 Add Guard dataset curation pipeline` added Spark-backed dataset curation.

The source policy is now:

- All existing Guard data, including legacy holdouts, is authorized as a curation source by user mandate.
- Former holdouts can enter a new training version only after source-hash re-splitting produces new validation and regression manifests.
- Existing benchmark suites must be replaced/versioned before comparisons if their old holdout sources are used for training.
- Source data is never overwritten; every output is a new immutable `DatasetVersion`.

Spark Guard location:

- `/home/daniel/forgeguard`
- Active data root inspected: `/home/daniel/forgeguard/project/data`

The first bad recursive Spark inventory created `227,137` null-record-count DB rows from individual evidence JSON objects. Those malformed local DB rows were purged. The importer now accepts only explicit manifest paths; it does not recurse through evidence stores.

Qwen Dataset Curator proposal:

- Proposal ID: `0afcb3f5-a097-4461-b890-14d350e9d8f6`
- Qwen identified heterogeneous schemas, weak/contradictory labels, provenance risks, and exact-holdout handling requirements.
- Its proposal is persisted as `DatasetCurationProposal` and a Model Studio artifact.

Spark materialized versions:

1. `guard_curated_v01`
   - `/home/daniel/forgeguard/artifex_dataset_versions/guard_curated_v01/`
   - Train `4,684`; validation `657`; regression `541`.

2. `guard_curated_v02` (current recommended corpus)
   - Train manifest: `/home/daniel/forgeguard/artifex_dataset_versions/guard_curated_v02/train.json`
   - Validation manifest: `/home/daniel/forgeguard/artifex_dataset_versions/guard_curated_v02/validation.json`
   - Regression manifest: `/home/daniel/forgeguard/artifex_dataset_versions/guard_curated_v02/regression.json`
   - Artifex DatasetVersion ID: `416fc9dd-c922-4581-8ed5-ca10a488d224`
   - Train: `3,030`
   - Validation: `437`
   - Regression: `332`
   - Train SHA-256: `d4c1a6accee0fac511641a24287cfdb469819fdd6d34af720c7349dc66b4803c`
   - Validation SHA-256: `3e3d5ad0d306bdd4a530fd2cf1d32bea030980af8d964d95c34b28237f6d5f`
   - Regression SHA-256: `a868c4221f291953a91cf5e03f8163b2fa4e134db5f21e369fbf7c75c2387f1c`
   - Source records read: `10,057`
   - Accepted after strict repair: `3,799`
   - Exact duplicates removed: `4,099`
   - Malformed/weak-schema records removed: `2,159`
   - Strict repair requires Solidity pragma, parseable JSON output with `findings`, and excludes `c4_invalid` records.

No Guard model training was started. The next required milestone is to deploy/sync Artifex state to Spark, register a new evaluation suite version using the v02 validation/regression manifests, run the fresh base-model baseline on Spark, then create the first actual training canary.