Inspiration

Payments today live outside of language. You describe what you want ("split dinner with Ana", "pay my teammate for the design"), then you leave that meaning behind and go click through a checkout. I wanted the opposite: PayPal as a word you can say, with a real transactional consequence on nodes in a network. The meaning itself moves value.

What it does

.wallet / PayPal adds PayPal to .me, a semantic language that runs on a deterministic, reactive kernel.

  • You (or an LLM) describe a payment in plain language.
  • The AI turns it into .me statements: who pays, who receives, how much, and how it's split.
  • The kernel derives every value in a live graph. Change one input and only the dependent nodes recompute, so you can see exactly why a number is what it is.
  • PayPal (sandbox) executes the operation, and the receipt is written back into the same graph.

Credentials never travel with the meaning. The app asks the namespace for them at me.keys.paypal. Secrets are derived and kept locally, and they are only ever shown as dots.

How we built it

  • .me (npm i this.me): the deterministic kernel. Expressions like me["="]("share", "total / people") become derivations in a dependency graph.
  • .GUI (this.gui): the interface layer, with the topbar shell, the identity QR (a cleaker.me namespace), and the LLMS chat component that puts the model in the loop.
  • PayPal Developer Platform (sandbox): creating and capturing orders from the derived values.
  • AI: an LLM reads the user's intent, writes .me statements and explains each transaction back in human terms. The kernel, not the model, decides the numbers, so the result is deterministic and auditable.

Challenges

  • Keeping the LLM honest. The model proposes and the kernel computes, so a hallucinated amount can't move money.
  • Separating identity, credentials (me.keys.paypal) and payment fields, so that no secret ever appears in a prompt, a page or a network call.
  • Making a reactive graph readable to a non-technical user in real time.

What we learned

When meaning is structured, payments become composable. A split, a recurring share or a team payout is just a few derivations. And an LLM is far more useful as a translator into a deterministic language than as the thing that decides.

What's next

Recurring and conditional payments, multi-party payouts across namespaces, and moving from sandbox to production.

Live demo: https://neurons-me.github.io/.wallet/paypal/ Code: https://github.com/neurons-me/.wallet

Built With

Share this project:

Updates

Submission history