Research that reads everything.
So you don't have to.
Flux pulls the 8-K at 4:02, the transcript at 5:30, every headline that crosses the wire and every unusual print on the options tape, then writes it up in plain English, with the model forecasts, the catalysts and the risks in one page.
Six streams, read in full, every session.
A person can read one 10-Q properly in an afternoon. Flux reads the filing, the call, the wire and the tape around it, and tells you which of the four actually changed the picture.
SEC filings & 8-Ks
10-K, 10-Q, 8-K, S-1 and insider Form 4s, parsed the moment they hit EDGAR. Flux diffs each filing against the last one and reports what changed, not what was restated.
Earnings calls & transcripts
Full transcripts summarized down to guidance changes, margin commentary and the questions management dodged, with the exact quote kept alongside the summary.
The news wire
Every headline touching your universe, read by the catalyst analyst and deduplicated, so one story syndicated forty times counts once.
The options tape
Sweeps, blocks, open-interest shifts, skew and gamma positioning. The tape often moves before the story does, and it shows up in the brief when it does.
Price & volume history
Years of daily and intraday history behind every claim, so "unusual volume" means a number against a base, and a level is one the tape has actually respected.
Your own portfolio
The part generic research can't do. Flux reads your positions, sizing, stops and correlation, and writes the section that's about your book, not the market's.
Six models. One report.
The write-up you read is the last step. Behind it, six specialised models each answer one narrow question, and the Reporter stitches their answers into something a human can act on.
Model names describe the class of technique used for each task. Forecasts are probabilistic estimates, not predictions, and no model output is a recommendation to buy or sell.
The desk files before you ask.
Four standing reports land on their own schedule, in the app, by push, or by email. Every time is configurable, and you can trigger any of them on demand.
Morning Brief
What moved overnight, what changed in your watchlist while the market was shut, and the day's calendar. Written before the bell so the first decision of the day isn't made blind.
- Overnight recap
- Watchlist deltas
- Model forecasts
- Today's catalysts
Midday Pulse
A short read on the session in progress: where breadth is, which of the morning's setups held, which failed, and any fresh filing or flow that landed since the open.
- Breadth & regime
- Setups holding
- New flow
- Intraday filings
Closing Wrap
The session scored. What worked, what didn't, what the after-hours tape is already saying, and the filings that dropped in the fifteen minutes after the close.
- Session P&L attribution
- After-hours moves
- Post-close 8-Ks
- Tomorrow's setup
Portfolio Review
The long look. Concentration, correlation, drawdown, position sizing against your own rules, and whether the edges you're trading still show up in the last twelve months of data.
- Concentration
- Correlation matrix
- Drawdown
- Rule adherence
Every claim gets a date it has to be right by.
Research that can’t be checked is just prose. Flux does not backtest — there is no historical replay engine here, and a number invented for a panel would be worse than none. What exists instead is a forward record: every prediction is filed with the date it resolves, and on that date it is graded against the tape whether it worked or not.
A call carries its resolve date at the moment it is filed, so it cannot be quietly re-dated once the tape disagrees with it.
The misses are written to the same record as the hits, against the analyst who made them.
Candidates that failed a gate stay on the record with the gate that stopped them, so what the desk threw out is as readable as what it took.
The desk grades a prediction only once it reaches its resolve date, so there is no curve to draw until enough calls have come due. When there is one it will be the record itself, not an illustration of one — which is why this panel is empty rather than filled with a shape.
No hypothetical curve · no backtest · no simulated track record
Counts read from the desk's own record. There is deliberately no win rate, Sharpe or drawdown here: the desk trades a simulated book, so any such figure would be hypothetical, and a prediction is only graded once it reaches its resolve date. Grades, including the misses, appear on the scoreboard as they come due.
What the research is, and isn't.
Is this investment advice?
No. Flux is research and automation software, not a broker-dealer and not an adviser. Reports summarise public information, model output and your own portfolio data. Nothing in them is a recommendation, and every decision, with its outcome, is yours.
Where does the underlying information come from?
Public filings, earnings transcripts, licensed market data and news feeds, plus the portfolio data you connect. Each claim in a report carries its source, so you can open the filing or the print behind it rather than taking the summary on faith.
Can the AI get it wrong?
Yes. Language models misread things, data feeds lag, and history is not a promise. That's exactly why every claim carries its source, and why every prediction carries the date it has to be right by and is graded against the tape when that date arrives — so you can check the work instead of trusting the tone.
Can I ask for a report on something specific?
Any time. Ask for a name, a sector, a theme or your own book in plain English and the Reporter assembles it on demand, using the same inputs as the scheduled briefs.
How current is a report?
Filings and wire items typically land in the research index within a minute of publication; scheduled reports are written at their configured time. Every report is stamped with the moment it was generated and the cutoff of the data behind it.
Does Flux backtest, and does its record mean I'll make money?
Flux does not backtest: there is no historical replay engine in this product, and any backtested figure you saw here would have been invented. The record it does keep is forward — dated predictions graded against the tape as they come due — and it is a record of a simulated book, so it is hypothetical too. It shows what the desk called and how those calls landed, not what your own fills, timing or risk tolerance would have produced, and not what happens next. Trading involves risk of loss, including loss of principal.
Tomorrow's brief is already being written.
Turn Flux on tonight and read the research over your first coffee instead of building it.