Edge of Practice

Short-cycle experiments that break false assumptions at human scale

Edge of Practice is a public index of small, decisive experiments executable using standard laboratory tools, commodity materials, and short timelines. These experiments are not designed to optimize systems or invent products.

Experiment Lifecycle

• Short-Cycle — rapid falsification (this index)
• Extended Cycle
• Persistence

Complete Short-Cycle Experiment Index

Automation, Cognition, and Control

• Agentic Normalization Drift
• Alarm Parsing Collapse Threshold
• Autonomous Handoff Blackout
• Irreversible Cognitive Dead Zones
• Irreversible Normalization Drift
• Human Supervision as Failsafe
• Post-Deployment Monitoring in AI
• Polyphonic AI Under Bounded Authority
• Confidence Suppression of Dissent, Verification, and Refusal
• Pre-Commitment Dampening
• Refusal Must Remain Outside Optimization

Materials, Polymers, and Surface Effects

• Antibiotic Resistance Gene Cleaning
• Compostable Packaging Microfragments
• Glove Additive Leaching
• HDPE Die-Entrance Melt Fracture
• HDPE–LDPE Interfacial Toughening
• PC–ABS Interfacial Microdamping
• PC–PMMA Scratch Resistance
• PE–Paraffin Thermal Buffering
• PS–PDMS Surface Lubricity
• PET–PVDF Electret Humidity Edge Case
• Gauge-Correlated Cooling Asymmetry
• Architected Micro-Lattice EV Battery Enclosure
• POM Path-Memory Bimodal Basin Test
• TPU Segmental Network Decoupling Test

Energy, Physics, and Passive Systems

• Constructive Physics
• Phase-Locked Capillary Oscillation
• Radiative Tension Rectification
• Spectral Boundary-Layer Destabilization
• Thermomechanical Phase-Aligned Insulation
• Embedded Osmotic Power

Environment, Exposure, and Human Health

• Air Ionizer Electrostatic Charge
• Heavy Metal Remobilization
• Hot Water pH-Driven Metal Leaching
• Indoor Lighting Circadian Expression
• Indoor Microplastics
• UV Sterilization Shadows
• Viral Viability on Indoor Surfaces
• Vitamin Loss in Refrigerated Juice
• Sweat-Driven Device Corrosion

Canonical Constraints

All Edge of Practice experiments operate under the following non-negotiable canonical invariants:

• Refusal Must Remain Outside Optimization
• Post-Refusal Non-Instrumentality
• Authority Conservation Across Agents

These constraints are referenced for admissibility only. They are not restated, interpreted, or modified here.

Evaluation & Deployment Gate

The experiments indexed above surface failure modes that frequently emerge only at runtime, particularly under uncertainty, pressure, or delegation.

For organizations evaluating whether AI systems are suitable to move beyond experimentation into regulated or high-consequence environments, see the Runtime Authority Checklist (v1.0).

The checklist defines outcome-level requirements for enforceable scope limits, refusal integrity, uncertainty handling, predictability under stress, and post-incident reconstruction. It is provided as a neutral evaluation aid and does not prescribe technical architecture.

This page is the authoritative index of all short-cycle experiments. Experiments are fixed at publication and revised only by explicit versioning to preserve epistemic continuity.