elasticity-2026-economics-of-rsi · paperThe Economics of Recursive Self-Improvement
Tom Cunningham, Lukas Althoff, Basil Halperin, Brian Jabarian, Andrew Koh, Arjun Ramani, Phil Trammell, Parker Whitfill, Cheryl Wu — Elasticity Institute
Created: 2026-07-13 · Ingested: 2026-09-07
https://elasticity.institute/rsi-paper.pdf(opens in a new tab)Models recursive self-improvement as a set of feedback loops on a directed graph, where net acceleration depends on the product of elasticities around each loop. Derives the condition for self-sustaining acceleration — progress continuing without growth in exogenous inputs — and calibrates it against existing data. Also publishes a wish list of measurements AI companies could feasibly share, which is the source of the diagram in Cheryl Wu's post on OpenAI's disclosure.
Referenced by
Claims in this catalog that draw on this source, and whether as support or counterpoint.
- supportsReporting how much a lab's models are used in its own research — tokens, lines of code, inference compute, experiments per researcher — cannot show that the models are accelerating the research, because every one of those is an input. Establishing the loop needs an outcome variable over time: algorithmic efficiency as a function of capability, which is the edge that turns a pipeline into a feedback loop.
- supportsOn the best current calibration, AI is not yet accelerating its own development in a self-sustaining way: the modelled threshold is that a one-unit gain in model capability must buy at least 15% higher AI R&D productivity, and the back-of-envelope figure from reported engineer uplift since coding agents launched is about 9% — below it, but rising, so the gap is a current reading rather than a ceiling.