Reproduce “An Operationalization of Opaque Serial Depth” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Nathan Sheffield, Alek Westover, Lukas Finnveden, Alexa Pan, Julian Stastny, Ryan Greenblatt

The authors define natural-language-rooted-node-separated (NLS) depth, a measure of the longest computational path that avoids qualifying human-understandable bottlenecks. A node qualifies when rooted in a pretraining prior, retains its output space, and is interpreted as the same kind of data. Their calculations put a hypothetical latent-reasoning design at 47.2 million, versus at most 17,000 for compared open-source chain-of-thought models; this measures architecture, not monitorability.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.