I agree this would make Lumo much more practical on mobile. If a response has already been submitted, it’s frustrating to lose the generation simply because the phone locks or you briefly switch to another app.

Ideally, the request should continue processing server-side after the app goes into the background, with the completed response available when the user returns. A notification could also let the user know when the response is ready.

It would be useful if this worked consistently for both short and longer prompts, while still giving users a way to cancel a request that’s taking too long. This feels more like an expected mobile-app behavior than an advanced feature, especially for longer AI responses.

Hopefully the Lumo team can consider it for a future update. https://coolmathgames.us.com/ can be a quick distraction while waiting for a response.

If I close the app while Lumo is generating a response, the answer is lost. No buffer, no way to retrieve it later. The reason is clear: LLM server is stateless, client is the persistence layer.

Other AI assistants (Claude, ChatGPT) solve this server-side – finish generating, client picks it up later. Lumo can't do that because of the no-logs policy.

But it could work encrypted:

1. LLM generates response
2. Encrypts it with my public key (they have it – it's public)
3. Stores encrypted blob in a sort of "Lumo inbox"
4. Deletes everything on its end – stays stateless
5. Client reconnects → fetches encrypted blob → decrypts locally → saves with zero-access encryption

Same principle as Proton Mail handling incoming emails. Infrastructure already exists.

No logs, no plaintext on the server, no persistent state. Just an encrypted drop waiting to be picked up.