A public challenge helped test Vercel’s agent sandbox

Transcript

Transcribed by machine (Whisper) from the episode’s audio, so a word may be off here and there.

1. 47:52

   And they're finding stuff that is just so bad. To get ahead of this problem, we put a million dollars towards a sandbox challenge. We told the world, Vercel Sandbox, we believe, is one of the most secure sandboxes, if not the most secure sandbox in the world. And to strengthen it further, we want you to poke holes. and we invited, I mean, you can imagine this, right? We invited the entire world to attack us.

2. 48:21

   Only one person with an army of agents was able to make progress and we're basically processing his contribution now and we're sharing it with industry peers. TBD. He and I are best buddies now and we're DMing on X as we speak. and for sure we'll be collaborating with him for a long time to come because I mentioned that the contribution is small but impactful. Great example. This gentleman, amazing cybersecurity brain.

3. 48:54

   This is not, you know, someone's child saying, hey, Claude, can you hack the sandbox? And, like, Claude did it, right? Like, this is very sophisticated cybersecurity skill plus an army of agents. And so what we're doing is because we invited people with this bounty, we now have, this is called the white hat sort of hacker, right? It's like a good Samaritan. We're now sharing this finding with, basically this is the challenge, everybody in the world.

4. 49:22

   Because this found a critical vulnerability in Linux. And almost the entire world runs in Linux. And so what Vercel is doing now is we're creating the sort of cybersecurity harnesses and the incentive systems such that we can find out first we can protect our infrastructure we can deploy lots and lots of tokens towards, you know, someone asked me on Exo like, what's going to happen in the future if all the human code that we've written to date is insecure, what's going to happen is like, look, I think in the short term there's going to be almost like a war between four good tokens and four bad tokens the reality is that you know, North Korea, China enemies, adversaries, bad people, good people we all have tokens now and the tokens I personally disagree with the whole like, let's ban open models because of the cybersecurity risk.

5. 50:14

   I think that's nonsense because the thing that is making these models good at cybersecurity is just that they're intelligent. And you would be basically prohibiting the access to intelligence. And this intelligence, we can deploy it in a defensive way, like we just did, right? And then help the entire industry. But I think there's going to be a rocky next couple of years because we're going to have to strengthen a lot of these foundations.

6. 50:38

   Now, really positive silver lining. I mentioned this weekend I worked on a proof system and a lot of my colleagues are researching this now. We can make the agents who are inexhaustible in their energy and capability basically create a kind of code that we can prove almost like mathematically is correct. And humans couldn't do this. Some humans, sorry, some humans in the audience, It's like a very small percentage of humans could do it.

7. 51:05

   But the code that will result from this looks more like the, I don't know if you saw that OpenAI proved Navier Stokes. And none of the mathematicians can understand this proof. I opened the GitHub repo and OpenAI can also not understand it. Like it's like so much code of this mathematical proof. And so I'm not saying that our machine code will necessarily look this way in the short term, but it's going to go more towards the mathematical certainty.

8. 51:37

   And so I'm going to be able to say, and by the way, Vercel has sort of been going down this path of strengthening our infrastructure with this more verifiable code for the past few years, but the surface is enormous. Most companies have not done this yet. Most of the code is still human-ridden. It's not correctly verified, and that creates risk for the world. but again it's it's gonna be uh i i think i just want to deliver a message of optimism i think it's all gonna be good and um everything's gonna be okay i swear it's gonna be okay yeah so we uh