Content pfp
Content
@
0 reply
0 recast
0 reaction

Venkatesh Rao ☀️ pfp
Venkatesh Rao ☀️
@vgr
So one of the moonshot things I think onchain tech needs is a native phoneOS. The horrendous and fragile flow of web app to wallet connect to wallet then back again to sign is not entirely crypto’s fault, and I suspect not solvable within conventional OSes…
8 replies
3 recasts
28 reactions

Venkatesh Rao ☀️ pfp
Venkatesh Rao ☀️
@vgr
It’s an artifact of physical constraints of phones and to some extent tablets. Small screen + touch = apps sharing screen is hard + modals suck By contrast desktop browser wallets seem to work well enough, modulo some race conditions bs between multiple wallets that are probably solvable within OS/browser API models
1 reply
0 recast
4 reactions

Venkatesh Rao ☀️ pfp
Venkatesh Rao ☀️
@vgr
Why do I think this is not solvable in traditional OS? Basically you need app-to-app decentralized trust model that’s not mediated by the device OS too much and a way for apps to inject composable components into each other’s UIs. Ie one app should be able to expose a negotiated plugin interface in another.
2 replies
0 recast
2 reactions

Venkatesh Rao ☀️ pfp
Venkatesh Rao ☀️
@vgr
My suspicion is that while this is borderline doable within Linux based OSes it is against commercial interests of Apple and perhaps Android too. Apple famously developed a secretive silo culture where teams don’t talk to each other except through management chain. Seems to have driven iOS design via Conway’s law
1 reply
0 recast
2 reactions

Venkatesh Rao ☀️ pfp
Venkatesh Rao ☀️
@vgr
An entity that builds such an OS would need some good open hardware to work with of course, but also a federated security model that mostly lets app-to-app protocols handle things with richer control than clumsy flow control passing. Basically an on-device ad hoc network of apps with rich message passing.
1 reply
0 recast
4 reactions

Venkatesh Rao ☀️ pfp
Venkatesh Rao ☀️
@vgr
The situation is actually surprisingly similar to how there are no good robot OSes. ROS does half the right thing of building a message passing layer on-device as the control plane for the robot, but as a guest process in Linux it can’t actually handle the real-time stuff that’s central to robotics. So it’s messy
2 replies
0 recast
2 reactions