OutLayer is the Verifiable OS for Agents

Captions as YouTube publishes them, usually machine-generated, so product and people names can be misheard.

1. 0:04

   AI agents are all the rage right now. But let me pose a question to you. An AI agent holds your keys, manages your money, and makes decisions on your behalf. How do you know the code actually did what it said? Well, Outlayer makes every agent execution cryptographically provable. Outlayer is the verifiable OS for agents. It's here on near protocol mainet agents today run on cloud providers and you're trusting things like AWS and Google cloud with

2. 0:37

   your agents keys data and logic and there's no way to verify if the code actually ran and you're kind of just living in this black box that is hopefully following instructions but you're not really sure. And even if it's following instructions, there's so many add-ons and things that maybe you try and skills that who knows what actually is happening. And Outlayer is here to help you understand better what's

3. 1:07

   happening with your agent behind the scenes. So let's take a look at what is being offered here at Out Layer. Uh if you go to the site outlayer.ai, AI uh you will see that everything is run on uh Intel TDX enclaves and the reason why that's important is because that's what a TE is and a trusted execution environment and you have this ability to now make sure that secrets are staying encrypted and every everything is

4. 1:40

   within control of what you expect to have happen and it's verifiably through add stations that are on near mainet. So what is out layer exactly? Well, it's a real operating system. It's not just a wrapper around some other kind of AI system or whatever else. It is actually a primitive that you can build on top of. You can use all of it or some of it or you know, however you see fit in what you are trying to do with your AI agents. Uh if you go from a glance here,

5. 2:15

   there's a lot of really great things that are happening uh from again encrypted file systems, hardware secrets, uh agent custody, access conditions all the way through and importantly there is at astation for all the different executions and activities that go on.

6. 2:35

   You can even take payments and do things like that. Uh it is again an opportunity for you as a builder to leverage these tools or you as a user just to be able to understand what's happening with your agent. Agents need infrastructure they can't trust. And typically what you see right now with the problem is that everything is in a blind trust. If you are feeding your agent information like data, secrets, uh managing funds through

7. 3:06

   wallets, things like that, uh you are trusting a cloud provider with that information and those that can be exposed in many different ways. where here on out layer they are providing a computation going through TEES they're using the Intel TDX hardware enclaves as part of that and so the operator cannot actually read your data uh they can't modify the results they can't run different code it is all within that trusted execution environment it's not a

8. 3:36

   pro promise they say it's a it's physics great and if you think about the you know the example here of you know Linux ver for applications and outlay or for agents it's a really nice marriage of putting together what this actually is as a product it is an operating system uh here you see uh on the screen this comparison of Linux versus out layer and it really breaks down the different

9. 4:08

   parts here of like how it actually serves as an operating system for your agent It is not. It's much different than just a typical application uh infrastructure here where you need to be more cognizant of again things like protecting secrets, making sure that privacy is baked into everything, stuff like that.

10. 4:34

    Uh everything here through out layer is based into a few different sections and we'll go through each section here now to kind of break down what is actually happening uh and available to you if you were to build using the outlayer uh OS system. First things first, the kernel uh this is a complete agent infrastructure. It is one single interface and they allow for all of this to happen. Uh it runs in a sandbox wasi process and every call is a hardware at

11. 5:12

    a uh attested and the reason why that's important again is because you know what is actually being run. You can verify what is actually happening. There are some key areas here. Blockchain access, encrypted file system, multi-chain wallets, uh verifiable randomness and built-in payments. What's great about this is that while this is built on top of near, you don't have to just use this for nearbased stuff because it is also multi-chain. Uh so that's a nice opportunity there as well.

12. 5:46

    Agent custody. So the key here is that you know you are giving keys and this doesn't have to just be wallet. This could just be again [snorts] API keys, all different kinds of keys. But specifically here again with wallets and dealing with payments and things like that. Um those keys can't be lost. Like it's not you're not worried about them being leaked or anything else like that. Um

13. 6:10

    you set policies, you set how they can be even spent. So so there's actually rules around like hey you can only budget so much money. Uh there's a really nice opportunity here as a builder or as just again someone who's running an agent to have the policies you're looking for built into this and making sure that it is following what you expect your agent to actually do which is uh a great thing.

14. 6:38

    Uh this breaks down again kind of just like what this looks like and for agents and kind of how this can be all laid out. There is end-toend encryption. Again, this goes back to I've said a few times here now about like how you are storing your secrets, how you're setting your permissions, everything else through uh it is again baked into this operating system for your agent.

15. 7:08

    The trust model here is really important and this is the big distinction between a cloud provider and something like this. So a typical cloud provider is not necessarily going to have these verifications built in. You are basically required to trust what is happening like what is available to you.

16. 7:30

    And so uh that is very different here than the outlayer method which is much more about you can verify exactly what's happening whether it's you know code running in the enclave um you know whether it's the execution was actually tested and onchain uh all that stuff is you can see all this breakdown through here and anyone can verify not just you as the user. So if you are running this as maybe um enterprise level whatever that might be you you it can be anyone

17. 8:05

    who can actually go and verify. So if I'm interacting with your service that you're providing with your agent I can still go back and verify this happened. This is built by builders for builders all the way through. uh everything is composable by default and that's why I started this off with that you can use all of this you can use parts of this there is no restrictions around how you

18. 8:33

    decide to build this into it uh best case scenarios yeah you put your whole entire agent and build everything on top of this but maybe you don't want to use all that yet you want to use parts of this that's all available to you from here so it's really nice that you have that opportunity and because again it is stateless by design you are really working on an assumption that you know you will not have crashes, you're not

19. 8:57

    going to have leaks, it's not going to just be zombie processes running. Um state is explicit. It's not just trapped in RAM. So you're not worried about like memory overflows and things like that. The beauty of this versus other things that you may have seen on the market so far is that it's fully running on mainet. Uh it is not a white paper. It is not just research. It is not just a hypothetical. Uh it is available today that you can actually go and build. Um

20. 9:29

    there are clear examples that have been built already using this system. Price oracles, uh near emails an email service, private token transfers, Dow voting, uh things, you know, all these different opportunities are all here today in production using Outlayer OS.

21. 9:52

    you can get started. There's there's some docs that are already set up. Uh you can go and get started there. Uh again, if you're building using something like Cloud Code, Codeex, whatever else, um you can just point them to the docs and kind of get started with Out Layer right away. Uh it is really easy to kind of just jump in and get going. Uh, and there's some nice opportunities there to be at the front

22. 10:17

    of the line here when it comes to verifiable agents. So, uh, one of the nice things here too is there's also a playground and you can actually try out some different things, uh, as well. You can also just again, like I said, you can go to the docs. We'll end on this note without layer.

23. 10:37

    The beauty of this I said a few times here is I can get started today. I can start building right away. I have opportunity of all the things around um privacy and and security and verifiable uh outcomes at my fingertips to build with today. The key thing here is that maybe your users are not ready for it, but as a developer, as someone who's using these agents, as somebody who's maybe building products around this, to

24. 11:12

    be able to integrate this from day one and not make your users worry about if this is being done, but you know that it's being done is a huge opportunity. And it's available to you right now live here on Nearet today. Check out outlayer.ai and get started now.