<?xml version="1.0" encoding="UTF-8"?><rss version="2.0"><channel><title>Irrevon — Research</title><description>Research notes from the Irrevon project: benchmark preregistration, prior art, and research integrity.</description><link>https://irrevon.vercel.app/</link><item><title>Preregistering a benchmark, including the result that would kill it</title><link>https://irrevon.vercel.app/research/preregistering-a-benchmark/</link><guid isPermaLink="true">https://irrevon.vercel.app/research/preregistering-a-benchmark/</guid><description>How IrrevonBench uses a two-stage freeze before live and confirmatory evidence, while disclosing synthetic development pilots and a falsification criterion that would declare the engine unnecessary.</description><pubDate>Tue, 21 Jul 2026 00:00:00 GMT</pubDate></item><item><title>Prior art: what we did not invent</title><link>https://irrevon.vercel.app/research/prior-art/</link><guid isPermaLink="true">https://irrevon.vercel.app/research/prior-art/</guid><description>Irrevon stands on decades of work — outboxes, idempotency, sagas, durable execution, agent-effect staging, replay, and reconciliation. This is the dated survey and its narrow inference.</description><pubDate>Tue, 21 Jul 2026 00:00:00 GMT</pubDate></item></channel></rss>