Research Promoted, Data Paused

Hi everyone,

A small but meaningful change went live on quarex.org today: Quarex Research was promoted to the top of the homepage hub (directly above Library, sharing the same gold color), and Quarex Data was removed from the homepage entirely.

Why Research moved up

If you look at the eight items orbiting the Library on the new homepage, seven of them are consumption — ways to read, browse, or look something up. News, Blog, Radio, Studies, Tags, Index, Politicians: all read-only experiences.

Research is different. Research is the only tool on the page that produces something. You give it a topic, it searches more than 48,000 indexed questions across over 1,010 Living Books, finds the most relevant sources, has AI write structured answers using those sources, and hands you a complete Research Package you can use.

That belongs at the top of the hierarchy. It's the answer to "what makes Quarex different from just searching Google?"

Why Data was paused

QuarexData (the election data API at quarexdata.org) is a real project — structured candidate data, partner-ready, useful — but it needs sustained attention I don't have right now. Election scrapes, the schema work, the partnership pitch to Democracy Works: all of it deserves real time, and putting it on the homepage when it's not actively being developed sends the wrong signal.

It's paused, not killed. Server, domain, scrapes, and database all still run. When I have a clean stretch of time to dedicate to it, it'll come back — probably with a partnership announcement attached.

How Research turns the Library into content

Here's the part that genuinely matters. Research is not just "ask a question, get an answer" — it's the production pipeline for almost everything Quarex publishes.

The flow:

1. Enter a research title — e.g., The History of the Apollo Program.
2. AI suggests starter tags from the Quarex controlled vocabulary (357 tags across three tiers) — a launching point, not a final answer. You then refine: uncheck what doesn't fit, and add specific names, places, or events directly into the search box. Proper nouns produce the sharpest results.
3. Full-text search pulls up to 40 topic questions across the entire library that match. You pick the 5–8 most relevant sources for your project.
4. Pick a research level — introductory, intermediate, or advanced. This controls depth and tone.
5. AI researches each topic via the Claude Sonnet API, grounded in the selected library sources, with citations.
6. Download the Research Package — a single file containing five output formats (A through E) plus the raw research content and the full source list.

The five output formats matter. Format A is a clean briefing. Format B is an essay structure. Format C is an outline. Format D is a script. Format E is a Quarex Blog Post — HTML output with Insight Panel annotations baked in, ready to drop straight into the blog with working deep-links to the source chapters.

The Research Package works with any AI — Claude, ChatGPT, Gemini, whichever you prefer. The package is platform-neutral; you pick the AI that fits your voice, budget, and workflow. Different AIs produce different essays from the same sources, and that's a feature, not a limitation.

What Research is actually used for

Once you see the pipeline, the use cases are obvious:

• Blog posts — Format E gives you a publishable HTML fragment in minutes, fully sourced with chapter-level deep links to the library.
• QuarexRadio scripts — Format D is the radio-broadcast format Lilly reads on air.
• Briefing documents for newsletters, presentations, internal memos — Format A.
• Educational content — outlines and study guides for PublicStudies curricula.
• Books and long-form work — the source list becomes a bibliography; the research content becomes raw material to expand on.

This is the leverage. Every Living Book added to the library compounds the value of every future Research run. The library is the substrate; Research is the multiplier.

Try it

Go to quarex.org/compose/, type a research title (something with a specific name or event works best), click "Suggest tags from this title," and see what happens. The whole pipeline takes about 90 seconds end to end, not including research time.

Then hand the downloaded Research Package file to your AI of choice — Claude, ChatGPT, Gemini, whichever you use — and answer the questions it asks (which format, any specific angle, etc.). The AI does the actual writing, grounded in the Quarex sources.

If you produce something useful with it — a blog post, a script, a briefing — I'd love to see it. That's the proof the model works.

— Peter