Uncompiled execution

Grok16 5.2.0 — dev runs at normal speed; compile is chamber-organized and ahead, not on every click.

Uncompiled chamber flow diagram
Python stays interpreted · native siblings wave-convert once into plane cache

Execution lanes

LaneDev behaviorspeed_demo @ 3s
PythonTrue interpreter (CPython / gpy-16 GrokVM)~843,829 ops/s
C / C++Chamber compile ahead — no line-by-line interpreter~87–95M ops/s after plane cache
CMakeConfigure + build once; bin reused~93M ops/s

Doctrine: data/field-exec-uncompiled-doctrine.json

Queen .launch chamber

  1. Folder mirror — compartment code in .launch chamber
  2. Trim excess — strip README, lockfiles, build cruft on singular plane
  3. Pick runner — Python entry stays interpreted; native sibling wave-converts
  4. Cache — fingerprinted launch-singular-plane/<hash>/plane-<stem>
  5. Run like a bin — execution timed; convert_ms excluded after cache hit

Module: Queen/lib/queen-launch-singular-field.py

Commands

# Uncompiled Python via chamber
QUEEN_LAUNCH_SINGULAR_FIELD=0 python3 Queen/lib/queen-launch-chamber.py run examples/speed-demo/speed-demo.launch

# Singular plane (compile ahead + cache)
python3 Queen/lib/queen-launch-singular-field.py run examples/speed-demo/

# Full versioned bench
SPEED_DEMO_TARGET_SEC=3 ./scripts/grok16-toolchain.sh exec-full-bench

Speed bench manual · CMake & linking