Quickstart: your first run

This is a start-to-finish walkthrough of a single research run — from typing your first question to reading, extending, and saving a cited brief. It takes a few minutes, and there is nothing to configure beforehand.

#Ask your first question

A run begins with one plain-language research question. You do not build a search query or pick sources — you describe what you actually want to know, the way you would ask a knowledgeable colleague, and the agent plans the rest of the work from there.

The best first questions are specific enough to have an answer in the literature but broad enough that the agent has room to gather evidence. A few examples that work well:

• Does SGLT2 inhibition slow the progression of chronic kidney disease in non-diabetic patients?
• What is the current evidence linking gut microbiome composition to response to checkpoint-inhibitor immunotherapy?
• Is there a causal role for TREM2 signaling in microglial clearance of amyloid-β?

Type your question and start the run. From here the agent works autonomously — you can watch it, but you do not have to steer it.

#Watch the agent work

Once the run starts, the agent works through the same transparent stages every time, and you can watch each one as it happens rather than staring at a spinner. The workflow runs in order:

1. Plans the question. It breaks your question into focused sub-questions and works through each in turn.
2. Searches the literature. It searches broad life-science literature — 40M+ papers — and weights top-tier, authoritative sources.
3. Reads full text. It reads entire papers, not just abstracts, to capture methods, results, and caveats.
4. Weighs the evidence. It types each finding and weighs supporting against contradictory evidence, surfacing contradictions and gaps.
5. Screens for reliability. It blocks retracted and hijacked-journal sources and flags corrections before that evidence reaches your conclusions.
6. Synthesizes a cited brief. It verifies every claim against full text and assembles the brief with a Trust Index.

You can open the reasoning behind any stage as it runs. A research run is thorough work, so longer or more contested questions take longer to gather and verify — the depth of the read is the point.

#Read your cited brief

When the run finishes, you get a fully worked brief that answers your question and each of its sub-questions — not a list of links. A few things to look at first:

• The Trust Index. Every brief carries a score from 0 to 100 and a trust band, backed by six facet checks. It tells you at a glance how much weight the evidence can bear.
• Numbered, verified citations. Every claim links to the exact source passage behind it. Citations verified against a source’s full text are marked distinctly from abstract-only ones — one click takes you from a claim to the original passage.
• Evidence types. Each finding is labeled causal, associational, synthesis, or preclinical, so you can tell a correlation from an established mechanism at a glance.
• Contradictions and gaps. Where the studies reviewed disagree, the brief surfaces the contradiction; where the literature has not answered something, it names the gap rather than papering over it.

Alongside the brief, a research notebook records the agent’s reasoning and its confidence in each finding, so you can trace how any conclusion was reached.

#Continue with follow-ups

A brief is not a dead end — it is a live research run you can keep working. Ask a follow-up in the same session and the agent continues from everything it has already gathered, so you can go deeper without starting over.

Good follow-ups tend to do one of a few things:

• Narrow the scope. “Limit this to randomized trials in adults.”
• Chase a thread. “Go deeper on the mechanism behind the second finding.”
• Probe a weakness. “What would resolve the contradiction you surfaced?” or “What evidence is missing to establish causation?”
• Turn evidence into next steps. Ask for testable hypotheses, experimental designs, or the mechanistic links behind a finding.

Each follow-up is answered with the same searching, reading, weighing, and citation verification as the first question — the standard of evidence does not drop just because you are iterating.

#Save or export the brief

When a brief is worth keeping, save the run so you can return to it later — it is preserved with its full citation trail and notebook, and you can reopen it as a live run and continue the conversation from where you left off.

Because every claim carries a numbered, verified citation, an exported brief travels with its evidence: colleagues can follow each finding back to the passage that supports it and rerun your reasoning rather than taking the conclusion on faith.

#Where to go next

That is a complete run — question, work, brief, follow-up, and save. From here, it is worth understanding the parts of a brief in more depth: how the agent types and weighs evidence, how citations are verified against full text, and what the six facets behind the Trust Index actually measure. Each deserves its own read, and each makes your next question a sharper one.