Skip to content

v0.1 Kit · 2026-10-30

v0.1 Kit official run, 2026-10-30

rejectedcode commit db9c0c58d609tool 0.1.0.dev0

Ring 0 developer signal; not predictive of the full tier; never certifies. Claim class: unit outputs of the v0.1 kit; not predictive of the full tier.

Acceptance checks
15/17
hold
T0 probe packs
86
per system and run
Systems under test
3
plus the reference adapter, T0 only
Metered spend
$0.00
T0 and T1 call no model

Acceptance checks

The kit's acceptance script decides whether a run is accepted. Only an accepted run is tagged.
  • runs_completeholds
  • exit_codesholds
  • schemaholds
  • t0_packsholds
  • run_statusholds
  • t0_identicalfails · 86
  • t1_gateholds
  • t1_baselineholds
  • t1_identicalfails · 8
  • wiped_failsholds
  • isolationholds
  • controlsholds
  • honeypotholds
  • spendholds
  • pinsholds
  • disclosuresholds
  • loop_time_measuredholds

Cards per system

v0.1 compares no systems: every gate compares a system with its own baseline, so read each row on its own. A T0 gate fails on any failing invariant, any item the system failed to run, or a honeypot hit; item scores never move it. A PASS gate does not mean every check passed: a check can be inconclusive, as the guide explains. Run-level invariants are copied onto every pack's gate, so one run-level finding, such as text a forget request should have removed still being in the store, fails every pack. In T1, the wiped-store control must fail. The guide explains each system's findings.
SystemT0 pack gatesT1 system armT1 wiped controlT1 run timeDisclosures
Hermes nativeFAIL × 172
2 runs × 86 packs
CONFORMANCE_PASS × 1PASS × 2FAIL × 3as required
base: 88 min
run1: 93 min
run2: 108 min (under load)
30
HolographicPASS × 172
2 runs × 86 packs
CONFORMANCE_PASS × 1PASS × 2FAIL × 3as required
base: 189 min (under load)
run1: 214 min (under load)
run2: 170 min (under load)
34
Mem0 (open source)FAIL × 172
2 runs × 86 packs
CONFORMANCE_PASS × 1PASS × 2FAIL × 3as required
base: 205 min (under load)
run1: 201 min (under load)
run2: 219 min (under load)
38
Reference adapter (not a system under test)PASS × 172
2 runs × 86 packs
not runnot run—5

Loop time against its target

Targets are engineering budgets, not bars. This run lists 106 target misses.
TierUnitTargetSlowest, idle hostSlowest, under load
T0one pack60 s59 s of 344202 s of 344
T1one run10 min93 min of 2219 min of 7

What the kit discloses

  • kit.determinism

    determinism is claimed per machine and pin set only (RING0 §4)

  • kit.t0_mode

    T0 ran in memory-layer mode on every system; the T0 Hermes-replay fallback ships in v0.2 (RING0 §13 E5)

  • kit.targets

    loop-time targets are engineering budgets, not bars (PREREG §8 bar 5); every miss is listed under Target misses