Research

Papers, notes, and benchmarks from the lab. Placeholders are honest — we don’t present editorial cards as publications.

Featured

ResearchAug 26, 2026 · 12 min read

Distributed reasoning for science: why RACoN is built for the scholarly record

How a distributed LLM changes what a research lab can verify, cite, and reproduce — and why it belongs next to the journal, not in place of it.

How to read this index

Each card links to a template article page. Replace with real papers when ready; the IA already mirrors openai.com — research is parallel to product, not buried under it.

Index

All research

ResearchAug 14, 2026

Mapping citation graphs at scale without losing provenance

A method for large-scale citation resolution that preserves source fidelity.

14 min read

ResearchAug 01, 2026

Evaluating LLMs on peer review: a benchmark and its limits

We test model assistance in review and show where human judgment still dominates.

11 min read

EngineeringJul 28, 2026

Field notes: distributed inference for long-context scientific documents

Early lessons from RACoN’s inference layer on 100k+ token papers.

9 min read

ResearchJul 16, 2026

Open access does not mean open-washing

Definitions, checks, and tooling we use to keep “open” honest.

6 min read