Sthira

Feature · intelligence

LLM Explainer

Ask anything. Get an answer. Get no trades.

At a glance

  • Ask why your portfolio drifted, what an expense ratio is, anything
  • The model has read-only access to your portfolio state
  • It cannot, by architecture, originate an order
  • Sources cited where applicable — no hallucinated numbers
  • Conversations stored privately; never used to train any model

What it does

The LLM in Sthira is a teacher. It can read your portfolio state, look at your goals, walk you through the math behind a rebalancing proposal, or explain what a TER is. What it cannot do — by architecture, not by policy promise — is place an order. The order path has no LLM in it.

This is a deliberate design choice, and we’ve written about it in some detail: see Why we don’t let the LLM trade for you on the studio site. Short version: a layman investor cannot give informed consent to a black-box decision-maker that operates faster than they can read. So we don’t ask them to.

The LLM is genuinely useful for everything around the trade: explaining, teaching, surfacing relevant context, and turning jargon into English. Just not for pulling the trigger.