01 · The engine
Atlas
The retrieval-first system that powers every research turn. Plans, searches, drafts, and surfaces every citation with a paragraph-level anchor.
Learn moreFor the complete documentation index, see llms.txt.Skip to main content
The AI research analyst for foreign-policy work. Country briefs, policy memos, and Congressional testimony — with a primary source on every claim.
Between March 2022 and October 2024, India abstained on each of the seven General Assembly resolutions concerning the situation in Ukraine on which it cast a vote, including the emergency special session resolution of 2 March 2022 and the 14 November 2022 resolution on humanitarian consequences.
The Ministry of External Affairs has consistently called for adherence to the Charter and for respect for the territorial integrity of states, without naming the Russian Federation as the author of the conduct at issue. The Permanent Representative’s explanation of vote framed the abstention as a continuation of India’s long-standing position favouring diplomacy and dialogue.
The bilateral track has, over the same period, expanded rather than contracted. The 9 December 2024 India–Russia joint statement reaffirmed the “special and privileged strategic partnership” without revisiting the question of Ukraine. The framing is consistent with the literature on non-alignment as updated for a multipolar order.
Built on the primary record
The retrieval-first operating system for foreign-policy work.
The model never writes from training-data alone. Every research turn fetches primary documents from the index first, then composes.
Recent shipments
Four shipped surfaces. Each does one thing very well, and they compose into a single research environment.
The retrieval-first system that powers every research turn. Plans, searches, drafts, and surfaces every citation with a paragraph-level anchor.
Learn moreEnd-to-end execution of complex questions. Runs parallel searches across UN, courts, ministries, and journals, verifies, and delivers a draft you can hand over — every claim sourced.
Learn moreDiplomatic intelligence, refreshed every morning. Numbered developments, deep dives, and a path back into research on every story.
Learn moreA sustained research workspace. Keep context, custom instructions, and prior threads in one place — the agent reads them on every new question.
Learn moreFAQ
Short answers. Long answers in the methodology paper.
Every claim links to a primary source. Each session produces a full audit trail of which documents informed which paragraph, with retrieval timestamps. The audit is exportable for donor reporting and grant-funded research.
ChatGPT is a general writing tool optimised for fluency. Model Diplomat is a research analyst, retrieval-first by architecture — it searches the institutional record (UN ODS, treaty series, IMF Article IVs, country reports, peer-reviewed journals) before it writes, and writes only from what it retrieved. Plus the tradecraft templates ChatGPT cannot reliably produce: DC-format policy memos, CFR-style backgrounders, Congressional testimony outlines.
The retrieval pipeline cannot fabricate a document that does not exist in the index. Citations resolve to actual documents at the paragraph cited. We are publishing a methodology paper with the false-citation rate on policy queries; until that is out, we describe the architecture rather than assert a percentage.
Yes. Design-partner pilots are free for ninety days. We ask for a thirty-minute feedback call every two weeks. No high-pressure conversion.
Drafts and queries are never used to train models. Your account is isolated by row-level security; data is hosted on EU and US infrastructure. SOC 2 audit, single-tenant deployment, and SSO/SAML are on the roadmap — happy to share the timeline on request.
It does not predict events. It does not produce policy recommendations from thin air. It does not cite a document that is not in its index, even when asked to. It is a research instrument, not a forecaster. The argument is yours; the apparatus is ours.
We’re in private beta and shaping the product with the analysts, researchers, and policy professionals who’ll use it. Tell us about the work you do — every conversation moves the roadmap.
Retrieval-first architecture · No training on your data · Built for foreign-policy work