Say what your company needs to watch. Pol hires the scouts, gives each one its own computer, and publishes one issue with every claim linked to where it came from.
This week, your scouts filed
Meridian
Issue 12· this Monday
From the editor
Pricing is the story. Northwind is now under us at the entry tier and Halcyon has gone usage-based in the same fortnight — that is a pattern, not two events.
Reported by Ada, competitive intel
Northwind entry tier now $39
Down from $49, effective immediately. No grandfathering announced.
northwind.example
Every claim carries the link your scout read it from — so you can use it in front of a customer.
Why it matters
You already knew. It led issue 12, on Monday, with the link.
Most competitive intelligence goes stale within weeks of the thing it describes, because a person maintains it and that person has a day job. Your scouts check every week, and the issue is written for the whole company rather than for whoever happened to set up the alert.
Your first week
Not a trial that waits for you to configure it. You type one sentence and a research function starts running on your company the same afternoon.
Minute one
You get a named roster, not a loading bar
Your sentence comes back as people: a competitive analyst, an ecosystem watcher, whoever else that job implies — each with a face, a beat and the brief it was hired against. You read who you got before any of them has filed a word.
Minutes 2–6
Each scout is built and goes to work
One planning pass per scout, in parallel, then each one's program runs in its own container — fetching pages, parsing changelogs, following what it finds. A first run takes one to four minutes. They are real programs on real machines from the first minute, not prompts waiting for you.
The same day
Their first filings land in one thread
Findings arrive inside the message — the headline, what matters, and how many sources it rests on — rather than as a link to a report somewhere else. This is also where you talk back to them.
Then
Issue one publishes
Numbered, dated, with an editor's note saying what the week amounts to and every claim carrying the link it came from. It publishes on the next hourly sweep, or the moment you press Publish now. It belongs to the company, not to whoever signed up.
Every week after
It keeps going without anyone remembering
Scouts file on their own schedule, anchored to your timezone. A new issue publishes when there is something new to say — and doesn't when there isn't. Nobody has to own it, chase it, or put it in a calendar.
About four minutes a week, and a colleague can read the same issue rather than asking you what was in it.
“Stop covering funding rounds under $50M.” That is stored against that scout and shapes every future issue — not just the next reply — and you're told at the time what was learned.
No feed lists to curate, no schedules to set, no keywords, and no analyst whose side project this quietly becomes. That is the whole difference between a research habit and a research function.
Your scouts
These are the scouts a dev-tools company tends to get. Yours come from your own sentence — a bond desk and a clinic end up with rosters that have nothing in common, because the roster is designed rather than chosen from a list.
What your named rivals actually shipped — changelogs, release feeds and pricing pages, not press releases. Day two leads with what changed, because a battlecard that repeats itself is the stale one.
Breaking changes, security advisories and version drops across the stack you build on, each with what it means for you rather than a link dump.
Comparable launches with their numbers showing, and the objections in their comment threads — your future objections, for free.
Your editor. It never researches — its whole job is reading what the others filed and deciding what leads. That is the one judgement no scout can make alone, because no scout can see the others’ work.
They sit in one thread, so you can reply to an issue and get an answer from the scout it’s actually for — one of them, once. Tell one to stop covering funding rounds under $50M and it holds onto that for every future issue, not just the next reply.
Why you can quote it
Asking a model to be careful is a wish. Each of these is a path the model cannot reach, a link carried out of a database, or an absence we store rather than paper over.
It returns a headline, a note and an order. The findings and their links are carried out of the database by code — there is no path from the model's response to an item on the page, so a hallucinated one cannot reach you. A test feeds it fabricated sections and asserts they vanish.
Citations are numbered down the margin of the issue, and the count at the bottom is tallied from the stored items rather than asserted. An issue you can't check is one you have to take on faith, which is what the $20-40k incumbents sell.
If the synthesis fails we publish your scouts' findings anyway, in roster order, and the page tells you there is no editor's note on this issue. Withholding real findings would be wrong; passing roster order off as editorial judgement would be worse.
Scouts are told to say when a week was thin rather than pad it, and an issue only publishes when something new has been filed. A newsletter that always finds three important things is telling you about its prompt, not your market.
Pricing
$0
$99/mo
Free while in beta.
Name your competitors and the stack you build on. Your scouts are hired in under a minute, and your first issue publishes as soon as they have read the week.