thinkwright — AI research & tooling by Brandon Huey

Research & Software Projects of Brandon Huey

Evidence-first analysis of AI infrastructure, policy, and practice, alongside the software that supports it.

01 / Research & Analysis

Research & Analysis

Long-form research briefs and continuously updated data products measuring how AI is built, governed, and adopted.

AI Reports

Research briefs on artificial intelligence policy, infrastructure, and industry dynamics: evidence tables, entity spotlights, and explicit assessments.

AI policyAI infrastructureEvidence-firstLLMHarness engineeringLong Horizon Agents

Interactive census

State of the Agent

Analyses of large-scale Claude Code agent behavior, with the full census running interactively in the browser.

Claude CodeAgentic codingAgent skillsAgent behavior

Data product

AI Delegation Curve ↗

A versioned index of AI decision-making across 9 domains, with inspectable scoring and methodology.

AI measurementPublic indexMethodology

Research workflow

Defensible Deep Research

Source-grounded research delegation: lower-cost workers read and draft while a frontier coordinator checks claims.

Open weightsDeep researchVerification

02 / AI Software

AI Software

Open-source software for specifying, evaluating, and operating autonomous coding agents.

◆ Continuity

A context continuity sidecar for Codex & Claude Code that watches work checkpoints at meaningful boundaries and harness hook events to handoff compact briefings to the next agent entering the workspace -- all without resuming session history.

Ambient contextHarness continuityClaude CodeCodexSidecar

◆ Evaluation