# SIM2-SIM5 Practical Suite

Status:

```text
SIM5_SIM_SERIES_CLOSURE_MIXED__PRACTICAL_PILOT_NEEDS_SCOPE_CONTROL__NO_PHYSICAL_CLAIM
```

SIM2-SIM5 tested the initial practical ITHKOR-SIM plan beyond SIM1:

- SIM2 checkpoint / record placement;
- SIM3 observable-preserving compression;
- SIM4 held-out adaptive update robustness;
- SIM5 aggregate readiness.

This is an engineering suite. It does not add a physical claim.

## Result

Raw outputs:

```text
experiments/217_sim2_to_sim5_practical_suite/raw/sim2_to_sim5_practical_suite_2026-06-23/
```

Key gates:

| Layer | Status | Counted Passes | Practical Reading |
| --- | --- | ---: | --- |
| SIM2 | `SIM2_CHECKPOINT_RECORD_PLACEMENT_MIXED__FINITE_STORAGE_BUDGET_BOUNDARY__NO_PHYSICAL_CLAIM` | 0/3 | Event-only checkpointing loses to uniform-time checkpointing. |
| SIM3 | `SIM3_OBSERVABLE_PRESERVING_COMPRESSION_MIXED__FINITE_DETAIL_BUDGET_BOUNDARY__NO_PHYSICAL_CLAIM` | 0/3 | Record-heavy detail placement is not a broad compression policy. |
| SIM4 | `SIM4_HELDOUT_WORLD_FAMILY_ROBUSTNESS_PASS__STRUCTURED_TOY_WORLDS__NO_PHYSICAL_CLAIM` | 3/3 | Adaptive update budget generalizes over the tested structured toy worlds. |

SIM4 has one pass-like stress row on `noise_drive`, so a real pilot must report
stress/noise rows separately rather than using them as support.

## Interpretation

SIM2-SIM5 does not close the full SIM program as ready. It says:

```text
Adaptive update budget is promising. Checkpoint placement and compression need
separate scoped policies or should stay baseline/uniform in the first real
simulation pilot.
```

## Boundary

The first broad checkpoint and compression policies failed or were mixed. Those
failures are retained; they are not repaired by SIM4.