Research, not execution: where Quant Compose stops
The workspace is a research product. It does not place trades, route orders or give investment advice, and every generated report says so.
It is worth stating plainly, in its own space rather than as a footnote: Quant Compose is a research and paper-mode workspace. It does not place trades. It does not route orders to a broker or an exchange. It does not paper-trade against a live brokerage connection. GET /api/health reports live_order_routing: false, and nothing anywhere in the codebase places an order. Everything the workspace produces — a backtest, a strategy specification, a risk report — is analysis, not execution.
What the workspace does instead
A strategy inside the workspace is a versioned, inspectable specification document: a set of rules describing entry, position sizing, stops, take-profit levels, a trailing stop and a cooldown period, built through a visual builder, a formula editor, or a Python escape hatch, all of which compile down to the same underlying object and the same simulated execution. That specification moves through a gated lifecycle — draft, research, backtested, validated, paper, approved, live, monitoring, paused, retired — and two of those stages carry hard requirements that cannot be bypassed. Reaching “validated” requires an actual validation record showing the result survives its own trial count. Reaching “approved” requires a recorded human decision naming the specific evidence it rests on. Neither of these gates can be set by an automated process, and neither can be reached by skipping the stage before it.
The backtest itself is an event-driven simulation with an explicit cost model — a default commission of five basis points and slippage of two basis points, applied on every executed change in exposure, with an optional short-borrow cost accrued per period for short positions. A no-trade band skips rebalances too small to justify their own commission cost, and every simulated fill records its slippage and commission explicitly, so a researcher can audit exactly what a strategy’s backtested performance assumed about the cost of trading it.
What “live” and “monitoring” mean here
The strategy lifecycle includes stages labelled “live” and “monitoring,” and it is worth being precise about what those labels mean in this product, because the words themselves invite an assumption the workspace does not fulfil. A strategy marked “live” in the lifecycle has moved through the gated approval process described above — it has not been connected to a broker, and no order has been placed on its behalf. The Strategy Monitor page is a monitoring and record-keeping surface over a strategy’s specification and its backtest results: it lists saved strategies, tracks which lifecycle stage each has reached, and reflects the backtest results recorded against it. It is not a connection to a live or paper brokerage account, and it does not receive fills, positions or account balances from anywhere outside the workspace itself.
This distinction is stated verbatim, twice, in every generated PDF report the workspace produces: once in the footer of every page — “research / paper mode - live order routing disabled” — and again in the method and limitations section as its own numbered point. It is not a disclaimer added once and then treated as covered; it is repeated everywhere a reader might reasonably form the opposite impression.
Why the boundary is drawn here, deliberately
A research and analytics workspace and an execution platform solve genuinely different problems, and blurring the two tends to make both worse. An execution platform has to be fast, has to interface reliably with brokers and exchanges, and has to handle the operational realities of real capital moving in real time — partial fills, rejected orders, margin calls. A research workspace has to be honest about uncertainty, has to make it easy to interrogate why a result looks the way it does, and has to resist the temptation to make an unproven idea look more finished than it is. Quant Compose is built for the second set of problems, deliberately, and does not attempt to also solve the first.
This also shapes how a result from the workspace should be used by anyone reading it. A validated strategy, with a passing degradation grade and a deflated Sharpe ratio that survives its own trial count, is evidence that an idea is worth taking seriously — not an instruction, and not a guarantee about what will happen with real capital behind it. Nothing in a workspace report is investment advice, and every report says so, next to every verdict it renders.
What this means for how the workspace is used
For a student or an independent researcher, this scope is close to ideal: the workspace exists to answer the question “is this idea actually good, once tested properly,” without the operational complexity or the real financial risk of connecting it to a live account. For an institutional user evaluating the product, the same scope is a statement of what due diligence should focus on — the quality of the research and validation the workspace produces, not an execution capability it has never claimed to offer and does not intend to build in its current form. Anyone using Quant Compose should come away from that use with a clearer, better-evidenced answer to a research question, and nothing more than that.