# HBF scientific completion report — v0.6.11

Every score below is tied to an acceptance test that executed. “Complete” means
complete for the defined internal software gate, not experimentally validated.

| Scope | Score | Evidence |
|---|---:|---|
| Phase 1 — scientific contract | 100/100 | Registry, constants, run contracts, typed metrics, maturity taxonomy |
| Phase 2 — verification and uncertainty | **100/100 (15/15)** | Independent numerics, cross-runtime check, convergence, uncertainty separation, exercised covariance propagation |
| Phase 3 — scoped coupled research kernels | **100/100 (21/21)** | Energy/particle balance, alpha slowing, distribution framework, radiation channels, equilibrium, canonical stability, browser integration |
| Phase 4 — validation readiness | **100/100 (8/8)** | Reproduction runner, schemas, hashes, fixtures, preregistration and sign-off templates |
| Phase 4 — independent validation | **0/100 (0/3)** | No accepted external reproduction, external benchmark, or institutional sign-off |

Core test inventory recorded in `scientific-status.json`: **225 passed, 0
failed across 11 suites**, plus **243 passed, 0 failed** across the separate
Virtual Lab regression sub-suites. The
unified scientific release gate records **56 passed, 0 failed**. The complete
external runner executes 13 command groups.

## What changed from the supplied 93/100 and 81/100 scores

### Phase 2: 14/15 → 15/15

The missing covariance gate now executes end to end. The public data artifact
contains pointwise statistical errors and a 3.4% normalization systematic but
no released correlation matrix. Following a disclosed IAEA-style construction,
the code builds

`C = diag(stat²) + outer(systematic, systematic)`

and propagates `C_y = J C Jᵀ` through the actual log–log interpolated
Maxwellian integral. Symmetry, positive semidefiniteness, finite-difference
Jacobian consistency, and a seeded multivariate-normal linear-response check
are tested. The artifact is permanently labelled
`assumption_derived_lower_bound_covariance`; it is not source covariance.

### Phase 3: 17/21 → 21/21

The four missing gates were closed with deliberately scoped implementations:

- **Kinetic distributions:** positive normalized Maxwellian and thermal-tail
  distributions, moments, and the full isotropic two-species fusion integral.
  The Maxwellian limit and species-swap symmetry are tested. No Fokker–Planck
  evolution or tail-sustaining power is claimed.
- **Radiation:** relativistic electron–ion plus electron–electron
  bremsstrahlung and optically thin Maxwell–Jüttner synchrotron emission. Net
  synchrotron loss abstains because absorption and transport are not present.
- **Stability:** the exact linear ideal-MHD magnetized Rayleigh–Taylor
  interface dispersion, tested against analytic limits. It is not a global
  device-stability calculation.
- **UI integration:** `/scientific-core/` runs tested browser adapters, displays
  the claim boundaries, plots the distribution, and prevents stale exports.

The existing Grad–Shafranov solver remains verified against the analytic
Solov'ev solution with second-order convergence. A FreeGS/CHEASE comparison is
still an external strengthening gate, not an internal completion gate.

## Phase 4: what is fully complete

The software and documentation handoff is complete at `/external-validation/`:

1. one-command full and quick runners;
2. machine-readable environment, command output, duration, source-tree digest,
   and artifact hashes;
3. run-result and evidence schemas;
4. frozen benchmark inputs and expected outputs;
5. preregistration and independent sign-off templates;
6. browser-side structural result precheck;
7. reproduction and issue-reporting instructions;
8. an explicit prohibition on automatic maturity promotion.

This is **Phase 4 validation readiness: 8/8**. It is the maximum that can be
completed by the project author.

## Why independent validation remains 0/3

An internal agent cannot become an independent institution, operate an
unaffiliated code or experiment, verify a human identity, or sign on behalf of
a reviewer. Raising these gates without external evidence would fabricate
validation. They remain:

- third-party reproduction: pending;
- external benchmark result: pending;
- independent institutional sign-off: pending.

## Reproduce

```bash
python3 external-validation/run_validation.py \
  --institution "YOUR INSTITUTION" \
  --operator "NAMED OPERATOR" \
  --output external-validation/validation-run.json
```

See `HBF_SCIENTIFIC_RESEARCH_BASIS_2026-08-13.md` for the source review and
`platform/scientific-status.json` for the machine-readable gate record.

## Non-claims

Nothing in this release demonstrates ignition, gain, net energy, net electric
power, reactor feasibility, device-wide stability, or experimental validation.
The internal phase scores apply only to the declared equations, datasets,
domains, and acceptance tests.