Straight answers.

A dashboard with your profile, your settings, and, if you are a QuantDesk client, your desk. One account across the ventures that use it, rather than a separate login for each.

No, and it says so on the page itself rather than in a footnote. Submitting it opens the dashboard prototype with sample data. No account is created, nothing is stored, and no password is transmitted anywhere.

Yes, on every page, and it follows your operating system by default rather than overriding it. The toggle in the dashboard pins it, and the choice is remembered in your browser rather than on a server.

No. Not because the permission is switched off, but because the desk API has no write endpoint in it at all. There is nothing to enable by accident and nothing to exploit if a key leaks.

Indicative pricing is published on the API page even though nothing is chargeable yet, so that the first people to use it are not surprised. It is priced per finished minute rather than per API call, because a call that fails should not cost anything.

No. They are separate services with separate storage and separate keys. The only thing they share is the shape of their authentication header.

It watches a small number of markets, and for each one it has a written rule for what a trade must look like before it is allowed to happen: how far away the exit sits, how much of the account may be at stake, how many positions may be open at once. Most of the time nothing qualifies and it does nothing. That is the normal state, not a fault — a system that always has a reason to trade is a system whose reasons are not doing any work.

Partly. A language model is used where judgement helps — reading the day's context and proposing a shape of trade. It is deliberately not used for anything that must be exactly right. It never decides position size, never computes risk, and never places an order. Its proposal is rebuilt from scratch by ordinary arithmetic, then has to pass the same risk checks as anything else, and is thrown away if it fails. The interesting engineering here is the fence around the model, not the model.

Backwards from the exit. The distance to the stop-loss is worked out first, and the position is then sized so that being wrong costs roughly the same amount whatever is being traded. A volatile instrument therefore gets a smaller position, not a larger loss. There is also a ceiling on any single trade, a ceiling on how many can be open together, and a daily loss limit that stops new trades for the rest of the session.

Several separate things, on the assumption that any one of them can fail. Trades are defined-risk where possible, so the worst case is known before entry rather than discovered afterwards. A daily loss limit halts new positions. A reconciler compares what the system believes it holds against what the broker actually reports, and flags any disagreement. And there is one rule that overrides every other: nothing is allowed to stop the management of an existing position. A halt stops new trades. It never stops a stop-loss.

Honestly: it varies by strategy, and that is the truthful answer rather than a modest one. Some parts have a clear rationale and behave as intended. Others have been measured, found wanting, and either shrunk or left switched off. No past result predicts a future one, no figure on this site is a promise, and anyone offering you a number without a date and a method attached is selling something. You will see your own results, in full, including the bad ones.

Market data — prices, volumes, volatility and, where options are involved, what the options market implies about the range ahead. Plus a curated library of trading and risk-management theory the system can consult, and a written record of its own past sessions, so a mistake has some chance of being made only once. It does not use private information, and it does not know anything about you beyond your own account.

Usually because the conditions it waits for are not present. Selling options when they are cheap, or buying a breakout that has already happened, are both ways to be busy and wrong. Patience is a position. If a strategy has been deliberately paused you will see that on your dashboard rather than having to infer it from silence.

Yes. Every trade, every rejected proposal and the rule that rejected it, and an end-of-day report are all on your own dashboard. The reasoning recorded at the time is kept, including where it later turned out to be wrong — which is the only version worth keeping.

No. This site describes software and services. Nothing on it is advice, a recommendation, or a solicitation to buy or sell any security. The material in the Inspire panel is other people's work, linked because it is useful, not because anyone was paid to link it.

Yes. Trading in securities carries a risk of loss, automation does not remove it, and rules followed faithfully can still lose. The risk disclosure says this at greater length and should be read before anything else.

Your own broker, throughout. QuantDesk runs on your account and cannot hold or withdraw funds from it. The desk page explains the arrangement in full.