Relatable. While I doubt I'd have been able to do all the frontend myself without an agent, it's still pretty early for LLMs + money.

It depends on where you're starting a thread, but assuming you're starting the thread in a view, then we prime the chat with:

- The user's configured financial literacy, education level, etc.

- The reasoning and actual results of the calculator as you see them

- Summary metrics from the space that the view is saved in

The last point is pretty critical. The idea is that you build your own context from the bottom up, exposing as much or as little as you want for the scenario you want to explore. Add the relevant assets/liabilities/income/expenses, and we have a little cash-basis accounting simulator that projects all that into a "scope" usable by the Views, as well as the chatbot ("Finn" http://finfam.app/finn).

Handing these metrics to the chat unlocks a lot of the magic of the chatbot as a "neutral explainer". Any creator can upload a spreadsheet, but as we've currently exposed it, Finn is meant to be moderate and just there to close any explanatory gaps that the creator may have left out, or not considered. If you're looking at a view about buying a second house, Finn's supposed to give you a gut check if you have particularly low cash reserves.

A lot of a financial advisor's time and role is filling gaps in your financial literacy, and a 24/7 chatbot is honestly just better at explaining the difference between a credit and debit card.