Ask the house.

Every weekday, before the cash open, the desk sheets the names it would size — rank, ticker, thesis, catalyst, conviction. Each rank is held against the last two weeks and the last three months, so a name has to earn its place against what already worked. If we would not size it, it is not on the sheet.

A dated catalyst and a structure you can size around it. No catalyst, no sheet — flow, tape prints, and chatter never put a name on the letter by themselves. The setup has to be clean enough to write down: what has to happen, when, and what the number or the event has to do.

Sizing color, not an order. The ranks tell you how the desk would weight the names against each other that morning, highest down. It is how we see it — the work, and the risk, stays yours.

Every rank carries its catalyst and its date, and the calendar keeps every dated catalyst in one place — earnings, events, conferences, index effective dates. A name stays on the sheet while the setup is live; when the catalyst passes or the setup breaks, it comes off. Every dated sheet stays in the archive, so the record is checkable.

The letter is the ranked sheet: the setups. The tape is the session’s options flow plus the catalyst note — what traded, where the paper went, the structures worth a look. Flow is lookup, never a standing reason to sheet a name; it rides the letter only when it is the day’s cleanest setup.

The bar before the event: what the number has to do, not a recap after it. Earnings writeups are bound before the print — the bogey, the setup, the watch. After the print, the desk writes what changed, not what happened.

As a lookup, not a letter. The window is the full sheet of index events — ticker, side, shares, value in $m, xADV, effective date — sortable and filterable, so you can see what is coming and what size it is. Index flow never sheets a name on its own.

The letter is on the desk before the open. Through the session, notes land as the desk finishes them — earnings prep, the catalyst note, flow reads. Nothing is auto-sheeted; a name only moves when a human on the desk moves it.

TMT and the trades around it — semiconductors, memory, networking, AI infrastructure, software, social media, plus special situations where the setup is clean. The coverage is a high-conviction list, not the market. If a name is not on the sheet or in the book, the desk does not have an edge there.

It comes off the sheet. The letter is re-ranked every weekday from a blank page, so yesterday’s rank 01 has to re-earn it. The archive keeps every dated sheet, so you can see exactly when a name entered, held, and left.

Paid seats and active trials can send one ticker from Research — Request a name. The desk answers with a house writeup, an event bogey, an options catalyst, or earnings prep — when there is an edge. Not every ticker, not on demand, and a request never puts a name on the letter.

The whole desk, for seven days. The letter, the writeups under it, the rooms, the tape, the calendar — one trial per email. Leave inside the seven days and there is no charge; stay and the plan you picked starts.

One seat is the whole product: the weekday letter, every writeup bound to it, and the rooms. Research is not an add-on and there is no tier above the seat. Month $50, quarter $128, year $450.

From Account, two clicks. During the trial there is no charge if you leave. After a paid period starts, cancellation stops the next invoice — it does not unwind a period already billed. The seat stays open until the period ends.

No. Original house work — a synthesis of circulating research, specialist-desk commentary, and verified corporate developments. Not a scrape of broker PDFs, not a reprint, and not a recommendation to buy or sell any security. Size notes are desk color, not an order.

Because the models get finance wrong most of the time — and wrong with total confidence. The Financial Times covered Saturn’s testing of 18 models across ChatGPT, Claude, Gemini, Copilot, and Grok on 121 financial questions: wrong 57% of the time on average, and 88% wrong on the complex multi-step questions — which is exactly what a trading question is. Invented rules, missed risk warnings, calculation errors, all delivered in the same authoritative tone as the right answers. Free models were worse still.

A summary is only as good as the judgment behind what gets kept and what gets cut, and the study says that judgment fails most of the time — worse the harder the question. The seat is the opposite trade: good data, a model that knows what it’s looking at, and a desk that ranks a small set of names it would actually size before the open, each bound to a dated catalyst and a writeup you can check. The machine gives you everything. The desk gives you the sheet — and says no far more often than it says yes.

No recaps, no news feed, no reprints. No macro tourism, no filler to fill the page. On a day with nothing clean, the sheet says so — the letter is short when the market gives it nothing to say.