Start from the decision, not the method.
You don't reach for research because you want a horizon scan. You reach for it because a plan is due, a competitor moved, or a meeting is on the calendar. Pick the moment first. Signals routes it to the scans that answer it.
Know the method you need already?
Browse scans by typeThe strategy deck needs external evidence under the slides instead of relying only on the team's own read.
Every assumption gets evidence with a source behind it. No slide rests on a claim you can't trace.
A new space is on the table, for entry, build-vs-buy, or an investment case, and last year's analyst report is already stale.
Technologies land sorted by maturity (deployed, piloted, announced), each placement carrying the public evidence behind it.
You're tracking a named set of rivals, and the intel lives in a Slack channel of screenshots that no one can quote.
Every move ships with the page, filing, or posting it came from, so you can quote it in a deck without verifying twice.
You're in a regulated or exposed category where finding out from the trade press a week late is the failure mode.
Every signal links to the primary publication, so a board escalation has evidence behind it.
A board meeting, a strategy session, a roadmap review, and you've got days, not weeks, to form a defensible point of view.
Model agreement shows when the read came through more than one lens.
You've got a shortlist of places in your head, and you need to pressure-test it against real conditions, not gut feel or whoever pitched loudest.
Every location's read is grounded in sources, the rules, costs, and conditions you can cite, not a relocation brochure or a gut feel.
You're writing a trends report, a keynote, or a thought-leadership piece, and every claim needs visible evidence.
Every signal carries the sources behind it, so your published point of view survives the "where did this come from?" question.
Questions
What does “verified” actually mean on a signal?
Every signal carries its evidence. During a scan we run a source-and-verify pass: each claim is checked against the sources it came from, and those sources travel with the signal so you can open them yourself. “Verified” means the conclusion is backed by evidence we can show you, not a model’s say-so.
Why run several models instead of one chatbot?
A single chatbot gives you one model’s view through one lens. Signals runs the same brief across multiple frontier models in parallel, each blind to the others, then merges and scores what they surface. Where the models converge you get a consensus signal; where one stands alone you get a more novel, potentially disruptive one. Their agreement is itself a quality measure: the result didn’t come from one model’s bias, it came through all of them.
What stops the AI from inventing a trend?
Two things. Signals are grounded in retrieved sources rather than generated from memory, and a separate fact-checking step re-reads each one against its sources before it lands in your research record. A claim with no source behind it does not ship as verified. In practice about 95% of signals come back verified, and the rest are flagged for review rather than quietly kept.
Can I see the sources behind a conclusion?
Yes. Sources are attached to every signal, and you can trace a radar or report back to the evidence behind it. The whole point is research your team can defend.
Can I limit it to peer-reviewed or high-quality sources?
Yes. Verification is configurable: require peer-reviewed sources only when rigor matters, or allow broader sources when you’re scanning for inspiration. Either way, every signal carries the sources it was checked against, so you can judge the evidence yourself.
Which AI models do you use?
A scan runs a multi-model pass rather than betting on one provider, routing through the Vercel AI Gateway to the major labs. We pick the model that does each step best (retrieval, verification, synthesis) instead of asking one model to do everything.
What is a scan?
A scan is a multi-model pass over a research theme. It gathers sources, runs source-and-verify on what it finds, and returns a radar of signals with a starter report. Each signal includes sources you can check.
What do I get back from a scan?
A radar of verified signals for the theme, the sources behind each one, and a starter report you can build on. If you run a free scan, that output becomes the first Session in your Workspace when you sign up.
How is the free scan different from a deep scan?
The free scan is a starter Workflow on one theme, enough to see the method and the citations for yourself. A deep scan inside a plan runs the full source-and-verify, briefing-assistant, and assessment passes, and you can run them on your own themes as often as your monthly budget allows.
Is there a free trial?
No trial, but every visitor gets one free scan: enter a research theme and we run a starter Workflow on us. The output becomes the first Session in your Workspace if you sign up.
More than one of these on your desk?
Start with the question that matters most.