Designing boundaries for an Opus 5.5 + Jev trading agent
The author breaks down his trading agent architecture: Opus does the research, Jev makes only constrained buy/hold/sell decisions, and the output is a proposal rather than permission to place an order.
people keep asking me for the prompt that makes an AI trading agent profitable. i think that’s the wrong prompt to ask for. the interesting part of my Opus 5.5 + Jev setup isn’t a sentence that tells a model to “find good trades.” it’s the boundary around the sentence: -> what can Opus research? -> what must it return? -> what does Jev need to see before anything can happen? a trading prompt should produce a proposal, not permission to place an order. i broke down that separation and how the two models work together in the article below ↓
This post quotes another post: x.com/lostinazeroy/status/2103786949580345507