The Scurry Lab

A personal R&D lab running on
human-AI teaming.

A career's worth of questions. Finally, somewhere to take them.

02Active Projects

Infrastructure Harness Layer

SCOPING

Research into a transparent middleware layer that intercepts standard API calls - invisible to calling agents. Hardware-aware harness configuration is the research direction.

PhaseResearch
Current MilestoneSequenced after Franklin Phase 1
Last UpdatedMay 8, 2026
ResearchInferenceHarnessOpen Source
03Articles
№ 004METHODOLOGY

Rigorous AI-Assisted Research

A Grounding Framework

Jun 2026~15 min read
№ 003METHODOLOGY

A Methodology for Human-AI Collaborative Authorship

Building a framework for an heterogeneous agent lab

May 2026~8 min read
04Framework
Harness Engineering Framework
A design vocabulary for the outer layer of AI deployment — four architectural layers, three profiles, two interactive tools. Franklin Phase 1.
Draft v0.1
InteractiveFramework Explorer4 layers · 3 profilesInteractiveAuthority × Autonomy MatrixRisk quadrants · Trust progressionProfileGlobal Harness Profile5 dimensionsProfileAgent Harness Profile5 axesProfileTask Harness Profile5 dimensions
Research Grounding Framework
A failure mode taxonomy and intervention library for AI-assisted research — nine failure modes, fifteen grounding interventions, mapped in a coverage matrix.
Pre-publication
InteractiveFM × GI Coverage Matrix9 failure modes · 15 interventionsReferenceFailure Mode Definitions9 modes · taxonomyReferenceGrounding Intervention Definitions15 interventions · 4 stagesReferenceSources & AttributionFM + GI lineageReferenceSource Validation LogValidation corpus