Research
Publications.
Empirical Benchmarks & Protocol Standards
Our research teams investigate the machine operability, integration friction, and runtime isolation of software infrastructure—measuring how autonomous AI coding agents discover, authenticate, and execute across modern developer platforms with minimal token tax.
Autonomous Infrastructure
Evaluating cloud code sandboxes, container isolation, and ephemeral compute environments for autonomous AI coding agents under real-world execution load.
Agent Readiness Standards
Developing empirical measurement standards (ARS 3.0) to evaluate how effectively AI agents navigate, authenticate, and execute against software APIs and documentation.
Context & Token Economics
Quantifying the hidden token tax and latency overhead imposed on agent operators by non-machine-readable developer surfaces and poorly structured docs.
Agent Security & Teardown
Auditing execution isolation, teardown verification, and preventing persistent cloud compute leaks during long-running multi-turn agent sessions.
Publications
Published datasets, test harnesses, and evaluations investigating developer infrastructure readiness. Both reports were corrected on 2026-10-12; see the Sandboxes errata and the ARS 2026 errata.
150 popular sites fetched as a declared research agent, an unlabelled script and a headless browser: 16.0% vs 18.7% vs 18.7% of homepages challenged or blocked — no significant difference. robots.txt targets AI training (23.6%) far more than user-initiated agents (12.1%).
Protocol-only probe of 3,000 of 26,365 remote endpoints: 56.9% work anonymously, 82.7% of auth-gated endpoints give a followable challenge, 13.3% speak MCP 2026-07-28, and 24.9% fail to report errors in-channel.
Pilot benchmark evaluating Modal, E2B, boat.dev, Vercel Sandbox, and Daytona across 119 front-door scans and 30 live agent trials (28/30 passed). Corrected 2026-10-12: prompts differed by provider, so the Integration Tax ranking is indicative only.
Evaluation of 100 developer platforms across 1,000 deterministic, simulated Pathfinder traversal scans (mean ARS 50.7, median 52, max 79). Corrected 2026-10-12: the data, not the PDF, is the source of truth.