Foundations

The work behind the product.

Papers, methods, and reproduction artifacts. Every benchmark number we publish traces back to something on this page.

Needlepath research

PAGE RULER results in full The run behind the headline: accuracy, delivered tokens, refusal rate, cost, latency PAGE Prompt injection benchmark Needlepath on AgentDojo, ETH Zurich's public benchmark for AI agent security PAGE FinanceBench results The fit-limits case: financial filings, where the whole document is usually the right packet GATED Benchmark deep-dive Full scoreboard, per-length results, method notes. Request access PDF Executive brief What it does and what it is worth PDF Technical paper Benchmark methods and results ZENODO Benchmark paper and data Zenodo record with per-item artifacts ZENODO Method report Selection method, measured CODE Benchmark repository Harness, scoring code, paired statistics

Platform foundations

PAGE Aethon Instant agent instantiation PAGE Tool Forge Tool compilation and schema protection CODE Agent Definition Language Open source standard for defining agents PDF From Agent Platforms to AI Execution Infrastructure Eight-page paper PDF Solution brief Governed AI execution

Need something that is not here? Request research access.

Talk to one of our founders

About fit, and how this applies to your business.

Prefer a calendar? Book a time