AI across your trading-strategy workflow.
Periscøpe's AI assistant helps at the four points where it's genuinely useful in algorithmic trading: writing strategy code, debugging a failed run, analyzing run results, and suggesting improvements. It is grounded in your code and your runs, and it proposes changes you approve.
Four ways the assistant earns its place.
Write strategy code
Describe a strategy in plain English and the assistant drafts Python you review, or scaffold from a working example. A starting point you own, not a black box.
Debug a failed run
Hand a failed run to the assistant. It works from the run's real issues, your strategy code, and the logs, so the fix is specific to what actually went wrong.
Analyze run results
Ask the assistant to review a run's results and explain what happened, grounded in that run's orders, fills, and metrics.
Improve a strategy
Ask for a review of your strategy and get concrete suggestions as diffs you can accept, reject, or edit.
Suggestions you can read, as diffs you approve.
The assistant returns changes as a diff in the strategy editor. You see exactly what it proposes and why, then accept, reject, or modify it. Conversations are kept per strategy, so the context builds as you work.
- Draft, review, and refine in Python
- Every change is a diff you approve
- Per-strategy conversation threads
- You stay in control of the code
One-click handoff from a run that failed.
When a run hits a user-fixable problem, Fix with AI and Debug Run carry the run's issues, your code, and the logs straight into the editor, opened at the exact version the run used. No copy-paste, no re-describing the bug.
- Fix with AI on a single issue
- Debug Run for all user errors
- Opens the editor at the run's version
- Grounded in real run context
What this is, and what it isn't.
A grounded development assistant.
A faster way to write, debug, analyze, and improve strategies you design, working from your real code and runs, with you approving every change.
An autopilot or a profit oracle.
It won't invent a profitable strategy, trade on its own, or guarantee results. The edge, and the responsibility, stay with you.
More on AI-assisted development.
AI in algorithmic trading: questions.
What can the AI assistant actually do? +
Four things, all grounded in your work: draft and write strategy code, debug a failed run from its real errors and logs, analyze a run's results, and review a strategy with concrete suggestions for improvement. Each result is something you review and apply, not something it does silently.
Does it write or change code on its own? +
No. Every change arrives as a diff in the editor that you accept, reject, or edit. The assistant proposes; you decide. It never edits your strategy or places trades on its own.
Is it grounded in my actual strategy and runs? +
Yes, and that's the point. When you debug a run, it works from that run's structured issues, the strategy code, and the logs. When you analyze results, it works from that run's data. The grounding is what makes its suggestions specific rather than generic.
Does it find profitable strategies for me? +
No, and you should distrust anything that claims to. It helps you build, debug, and refine strategies you design. It does not generate guaranteed or profitable strategies, and nothing here is investment advice.
How does the handoff from a run work? +
From a failed run, Fix with AI sends a single issue plus your code and config, and Debug Run sends all user-fixable issues. Both open the strategy editor at the exact version the run used, with the context already attached.
Build with an assistant, not an autopilot.
We onboard in focused cohorts. Tell us what you trade and how you work. Live access is rolled out to select users on request.