cyberelf.INDEPENDENT SYSTEMS LABTHE LAB / EXPLORATION IN PUBLIC
These experiments explore different kinds of systems. They stand outside the AI release chain and keep their own evidence boundaries.
A visual signal workstation for patching typed device racks across color, depth, motion, masks, normals, confidence, and stereo. Undo project edits, reject incomplete imports without losing your rack, and render shared shaders on the client GPU.
Requires WebGL2. Browser operation does not imply feature parity with the desktop Digital Vision Workstation.
Open the workstation ↗
A native PyTorch petri dish with a complete workbench, reproducible sessions, resumable genome evolution, and bitwise archive replay.
The image is an original concept illustration. The simulator uses an untrained update rule and a structured growth prior. See the actual workbench and replay evidence →
A search rig below model architecture: memory, consolidation, context routing, trust gates, and contradiction repair across replayable lifetimes.
Training, selection, and final testing are now separated. The corrected 48-trial run did not beat its baseline. Read the result and run the rig →
JUSTIN GARRINGER / FOUNDER
Cyberelf Labs is a founder-led systems lab in Waterloo, Iowa. We work between research and product, building runnable systems and the instruments needed to judge them.