A daily brief on your field, where every line can be traced back to what it was read from.
Your agent drains the sources you name every morning, scores what it finds against a hard bar, and writes a short brief of what actually moved. A quiet day is published as a quiet day. Every movement cites the items behind it, and every chart cites the file it was fetched from, so you can check any of it.
What it does
- Drain every source over the published window.
- Score each item for whether it moved the field.
- Write a short brief citing the items behind each line.
What you get
- A daily brief of what actually moved, or a quiet day said plainly.
- Every movement traced back to the items it was built from.
- A scoreboard of your team's dated, falsifiable claims.
What it needs
- The field you want watched
- What matters most within it
- The arXiv, OpenAlex and feed sources to drain
Every weekday morning by default, in your time zone.
Nothing goes out until you approve it, unless you say otherwise.
Who runs it
The brief lands in the room your team already reads in, and the replies to it become the scoreboard. Nobody has to keep a separate list of who said what.
- One room, one brief: The wire posts where the team already talks. Anyone who should read it is in that chat, and the board is one link away rather than a forwarded PDF.
- Claims are filed, not remembered: A falsifiable claim in the replies goes on the scoreboard with its confidence and the date it settles. A claim missing either is refused out loud, naming the person and the reason.
- Anyone can check a line: Every movement links the items it was built from, and every chart point carries the URL, the sha256 of the bytes and the fetch time.
What stays yours
- Your beat and your lens: The field, and what you want weighted within it. Both go into every prompt the desk reads, so the brief is about your field rather than a generic feed.
- Nothing goes out unapproved: The publish gate is a real decision in the app. When you name a person or an audience, only they can answer it. Turn it off and the wire posts on its own.
- Every item kept, in your workspace: Signals, verdicts, briefs and predictions live in the wire's own store. A duplicate is marked, never deleted, which is why the funnel numbers add up.
Questions
- What is the difference between a quiet day and a partial read?
- A quiet day means the wire finished reading and nothing cleared the bar. A partial read means it did not finish. The board labels them differently and never calls one the other.
- Why do most items score low?
- Because the score asks whether an item moved the field, not whether it exists. A real, competent paper that restates known results is correctly scored low and still filed.
- Does my field have to be on arXiv?
- No. Leave the categories blank and the wire reads OpenAlex and whatever RSS or Atom feeds you point it at: trade press, lab blogs, regulator bulletins, a newsletter.
- Can I plot someone else's dataset behind the brief?
- Yes. Give it a label, a CSV URL and the two columns. Every point is stamped with the URL, the sha256 of the bytes and the fetch time, so the chart can be checked against the file.
- Why was my prediction refused?
- Because it had no stated confidence, or no date it can be settled on. The desk says so out loud rather than inventing either. A scoreboard whose entries were defaulted is not a scoreboard.
- Can I change the sources later?
- Yes. The categories, the search and the feeds are settings, not code, and the next run uses them.
Read your field in five minutes, and check any line of it.
Install it, name the beat, and the first read runs at boot.
Browse all Bolter blueprints