System 02 · Market Briefing In production · daily
Morning News
A newsroom of agents that writes the market brief every morning — and a bench of editors with the power to kill the front page. Nothing ships on eloquence alone.
Why it exists
A plausible brief is worse than no brief.
Every market morning starts the same way: too many headlines, too little time, and a decision bench waiting. Language models can summarize the flood in seconds — but a summary that sounds right and a summary that is right are different products, and the gap between them is where money gets lost.
Morning News was forged on one conviction: verification belongs inside the pipeline, not on the reader's shoulders. So it is built like a newsroom with an old-fashioned power structure — writers write, editors judge, and the editors can kill the front page.
How it works
Three floors. Writers write, editors judge, the gates decide.
4 beat scouts ∥ · 1 author · 1 sources gate
Four scouts sweep their beats blind — rates & central banks, currencies, equities & credit, geopolitics — none sees another's copy. One author merges, scores and ranks the sweep into a ten-story front page; the sources gate then re-verifies every headline against its live source before anything moves downstream.
BuysCoverage with triage — the desk reads ten ranked stories, not ten thousand headlines, and nothing important slips through unranked.
3 researchers ∥ · 2 adversarial auditors · 1 narrator
The top three stories escalate. Researchers go deep in parallel; then two editors attack from different angles — one audits the reasoning line by line, the other hunts confirmation bias, challenging the house lean on purpose. What survives renders as interactive dashboards plus a plain-language read.
BuysDepth exactly where it pays — real research on the three stories that could move a portfolio, instead of shallow notes on fifty.
1 author · the day's hardest question · full gate sequence
Convened only when a story crosses a severity line. The output is a decision-grade written analysis — scenarios with probabilities, transmission channels, and an explicit what-would-change-our-mind section. Most mornings it never runs; that is the point.
BuysWhen it matters most, the desk holds a memo it can act on — not a headline it has to interpret.
Every tier ships through the same discipline — sources · reasoning · render — and rejected work loops back for a targeted rewrite, never a quiet pass. A fast morning runs about ten agents; a full edition, closer to two dozen.
The front page, up close
What lands on the desk.
A fragment, redrawn with mock stories — layout and scoring discipline are the real thing; every live edition carries its sources.
House rules
The discipline that makes it a system.
The agent that writes a section never audits it. Fresh eyes, every gate, every day.
Every statement traces to a checkable source. If the source can't be reached, the claim doesn't run.
When data is missing or a feed is stale, the brief says so — plainly, where it would have been. Silence is reported as silence.