# Actual uniform5 time/space pilot with qualified local Hessian assembly

Frozen before launch,2026-11-13. Full original C1 scope unchanged. Original
uniform20->10 at10us remains FAIL49.56%dh/76.12%motion;10/50/200us horizons,
physical normal flux/traction,exchange coefficient/new thin-element ratio,
global-fringe/heldout experiments still required. This pilot ONLY reaches10us.

Use separate tf_local_assembly_bench executable, ORIGINAL physical equations,
all energy terms/geometry/metric/H1 backend,tolerances and SI parameters.
Local Hessian assembly qualified on4full matrices/56new unit checks and
3actual controls162steps+4independent saved scalar/mechanical audits. Numerical
output equivalence was preregistered and passed; baseline actual data not rerun.
This is mathematically equivalent assembly, not the same executable as original.
No geometry-candidate dispatch,force corrections,C2kinetics or parameter fit.

New uniform5,N2112,domain320x160um,physicalcut240um,radius500um,SDS6mM,
insoluble,initial6um,no gravity/DLVO,variational preparation with released trace.
Run T10us atdt.0625/.03125us sequentially(160/320steps),observer on.
Save first step plus every1.25us(save_stride20/40); extra observations do not
change equations and original10us acceptance remains unchanged. More frequent
snapshots support independent physical-cut diagnostics without trajectory replay.

Before launch verify candidate code/binary hashes,ALL prior control raw hashes,
operator/unit/audit completion and original uniform10 references at BOTHdt.
No previous successful long trajectory is rerun. Existing guards reused through
exact hashes, not presented as new tests. Freeze full source/parameters/commands.

Require complete,failures0,N2112,ALL accepted force ratios<=1.000001,
global/regional V/S<=1e-10,energy-budget excess<=1e-7,prep pressure<=1e-8.
Independent SI/GEOS scalar+mechanical audits on immutable terminal-case copies,
unchanged existing formulas/tolerances; no auditing live native files.
Original moving-cell-overlap dh/displacement at10us:TIME<=2%,SPACE<=10%.
Compare uniform10->5 at EACH matchingdt;uniform5time pair qualified first.
Record supplemental common1.25us samples separately, never replace original
10/50/200us gates or claim10us covers the complete protocol. References are
not continuum truth. Retain all earlier failures and unresolved physical gaps.

Measured candidate5um starts~54s native timer. Engineering10us estimates with
1.3margin~3.14/6.25h,not controlled cost predictions. Set ACTIVE30min heartbeat
monitor BEFORE first Popen. Fresh native-tool authorization required by runner.
Parent also records own live native-handle/progress checks every1800s into
monitor JSONL. Record child/parent PID,full arguments,start/terminal UTC and
live process identity. Monitoring is read-only,quiet during normal progress;
notify completion/failure/action required. Never restart from a stale file,
observation timeout or transient failed poll. Native timeout is separate.

Hard percase wall caps18000/28000s,total native46000s(~12.8h). If first native
case fails,skip second and preserve all partial files; no retry. Entire budget
includes preparation/IO inside child runtime. No automatic horizon extension.
Record original full state and source even if budget expires,not completion.
After this batch terminal and reported,stop ONLY its monitor,not the C1 goal.

All native stdout/raw bytes,monitor checks,source/binary hashes and results
preserved. Frames>5MB local only; bounded ZIP chunks for exact reconstruction.
Scientific PNG+vectorPDF white Nature red/blue; update all4research entry docs
and own VariDrainage page as terminal results become available. No completed
full-time/space/C1 claim until actual evidence passes original requirements.