Hacker News new | ask | show | jobs
by gonehome 1402 days ago
Urbit is trying to build the tools that enable this (in part by fixing the ad driven engagement incentives that lead to centralization and the current state).

One interesting bit is if you’re making vegetables when everyone else is giving away heroin it’s not enough to just to make great vegetables, you really need to offer something that can’t exist outside of what you’ve built because your core technology does something different.

I think Urbit’s distribution and handling of auth could be that for distributed DevEx and building collaborative apps in a way that’s way simpler than on the current web stack. It’s not quite there yet but there’s a path to this reality and success is among the potential outcomes.
1 comments
elefanten 1402 days ago
Do you have any recommended introduction to Urbit that touches on these possibilities in particular?
link
gonehome 1402 days ago
Which particular ones? The DevEx auth stuff or the incentive fixing bit?

The ID model is probably where I'd start since it's what fixes spam/moderation problem in a way that can actually work without centralization. When IDs have a small, but non-zero cost spam becomes uneconomical and it's easy to block an ID. IDs also accrue pseudonymous reputation. From there you can start to fix a lot of the other stuff.

The DevEx bit also heavily depends on that. When you build the OS to handle IDs you move the abstraction up the stack. Modern operating systems don't do this in part because they were built before the web. As a result modern operating systems are largely just machines user's use to open a web browser and every centralized web application has to rebuild their entire auth stack and all of the collaboration tooling. They're incentivized to do this and be incompatible with everything else because centralization tends to lead to ad-driven business models and all the incentive problems associated with that.

If the OS handled IDs and collaboration you could just build your app, distribute it to the network and rely on built in OS libraries to do the complex work relying on guarantees from the OS itself. The users wouldn't need accounts on a bunch of centralized services.

People have tried to do this on the unix stack, but it fails for a few reasons[0][1]. It's not impossible to build something else that could work though and rethinking the stack from first principles leads to something that could work.

Imagine if all linux users could just dm each other because the OS itself handled encrypted communication between IDs. You could build and distribute linux apps without a mess of complexity just by publishing it to your node. You get the capability to build really collaborative applications out of the box because of a lot of these guarantees without having to give up control. There's a lot more possible than this in this model - I think it's a path to escape the local maximum we've been trapped in and get closer to the web the 90s cypherpunks imagined (and the personal computer hackers before them).

[0]: http://moronlab.blogspot.com/2010/01/urbit-functional-progra...

[1]: https://urbit.org/overview

link