I don’t view Breez as self custodial. They appear to manage liquidity by being a “man in the middle” for payments. It seems like they don’t custody payments for very long, but for a brief period they take possession of your funds.

I think it’s great that they do so because self-custodial Lightning is difficult to manage. I just don’t think they should say they are self-custodial when they aren’t 100% self-custodial.

I maintain that we’re moving now into a custodial world for most people. We’ll need to limit the custodial touch points and risks through smart technology (like Breez) and geo-distributed federations. https://image.nostr.build/3c5de55e420a17323c36cae0bf0520ac9f65234c77c851145f6099a5a47f140c.jpg  

 We intercept the payment, we don't have access to the user funds. You're also confusing trust and custody.  

 You only don’t have access to user funds because you choose not to access them, correct? 

It seems from your knowledge base that your software could change the payment details to access funds if you wanted. 

Anyway, sounds like a cool and useful solution but my point is that if you have control over funds it’s not self custodial. Perhaps “trust-minimized” would be a more accurate term. 

I haven’t dug into your source code so I might be totally wrong about whether your software has control over payment funds. Please let me know if I’m wrong about that. 

 Technically, any lightning node can intercept htlcs, it they can't access the funds because they don't have the preimage. So it's incorrect. There's an optional trust with using zero-conf channels (post interception). If users are choosing to use zero-conf channels, they trust the LSP not to double spend till confirmation. However, this is just one option of the LSP model.