Awesome Jev: 100+ POWERFUL Jev AI Use Cases!

Julian Goldie (0:00)

So today, we're going to be talking about some of the best use cases for using Jev. And this is from an open source gear project called Awesome Jev. There's actually loads of different public projects like this collected, but this one actually looked pretty cool. So you can see here, for example, there's over 100 different use cases for Jev, and these all link back to like open source projects so you can check out. So for example, we've got 24 here, 22 here, 20 here, 31 here, and a bunch of others as well. So this is pretty interesting. Now, if you're wondering, okay, why is this important? Well, Jev basically is very good for making decisions, which means that when you're using AI agents, it's essentially going to make decisions way faster, which means that you don't need to use like a big, heavy frontier slow model. So that makes your AI agents cheaper and also faster, because you can use Jev for most of the decision making. So let's say, for example, you were scoring like thousands of leads. Well, you could use Jev to do most of that. It'd be way faster than using something like CloudSonic. So let's have a look at some of the best ways to use this. Now, just as a fun example, you can see some of the game examples right here. So some people have been using this for, for example, like playing games automatically, because it's really good at computer use. So the way this does, the way this works essentially, is that it can control what's happening right here. You can actually run this experiment yourself by giving the architecture to your AI agent and using the setup, as you can see right here, as an example. There's another one, for example, where Jev actually plays Pokemon, right? So it can read Pokemon Red as a text and then answer type questions, which you can see right here, right? It doesn't have any vision. It can't actually see the game or anything like that. But what it can do is you can actually read the game's state as text and then answer type questions each turn. But because it's moving so fast, it can basically work straight away, and it can basically play that game automatically, as you can see right here. Pretty amazing. And again, you can run this yourself, so you can see how it works step by step and then run it yourself. Now, as well as for computer use, let me show you another example here in action. So we actually tested out this voice browser here, which you can see. So it can control Chromium. So we've got the voice browser on the left hand side, which is controlled with my mic, and then it can actually navigate over here on Chromium. So for example, for making very fast decisions, if you want to see this in real time, go to Google, open up Google, and you see it navigates straight away right there.

Really fast.

You just have to be clear with the instructions. So like go to Google, not very good, but open up Google, yeah, that works straight away. So you have to make sure that when you're using it as well, the decisions that it's making are plugged in, and if you're ever sure how to apply this to your own projects, one of the best ways you can do this is you can actually, for example, take the documentation from Jev AI and then go over to your AI agent and say, hey, how can we use this for this, et cetera, right? So for example, here I can say, what are some cool ways that I could use Jev for stuff I actually do day to day? Give me some example use cases. So we can ask it right there, and then it will actually start giving us some examples because Claude already understands us. We could do the same inside Chat GPT as well, but we'll just get it to come up some ideas right here. So for example, like triaging, that could be like leads that come to you, for example. You've got SEO tasks. Like for us, we have a lot of keywords to sort through, a lot of data inside Google Search Console, and we can actually use Google Search Console and the data inside there with Jev, and then Jev can actually look through the searches, make decisions based on that.

So for example, we could categorize the keywords from our Search Console into like which ones have buyer intent, which ones are informational, and which ones don't make sense. The same for example here inside the AI Profile Boarding Community. So you see how we've got all these different personalized examples, like even our agent operating system right here. We could use it for a router, routing between the best models or switching between local, cheap, and frontier levels depending on what we actually need. We could even, for example, use a router where it checks the action, and if it's like, okay, this probably needs a human in the loop, okay, check it before it goes out. And there's all sorts of cool things here. So if we have a look through this list as well, we've got a bunch for routing. So we've been through all the game examples. Let's have a look at classification examples right here. I think this just helps you train your brain on ways that you can use it, because it's quite a new concept, and a lot of people are not sure what Jev is or how to use it, etc. I've shown you a few examples already, but you can get even more inspiration from all this sort of stuff. So, for example, we've got an email intent workflow, where it can basically route your inbound mail depending on where it needs to go and what needs to be done with that. So it could look through the list and be like, okay, is it an invoice? Okay, it goes over here. Is it general? Okay, it goes to the general inbox. And Jev can basically automatically detect the intent of that email coming through to you. So that's the whole workflow right here. If we have a look at this example for the Pi Jev router, you can actually use it with your coding agents. So with this, Jev can actually choose a model and reason in effort for Pi, and then keep both fixed for the session. So this will basically automatically load the skill and set the model depending on what it needs. And you can have multiple different choices that Jev is trained on. So it could be like auto, high, low, omitted, right? And for example, if it's on auto, Jev will choose the lowest effort. It judges sufficient. If it's on high, it will force the level based on the model's capabilities. So it can basically automatically load the best model based on what you need to do. So there's so many different ways you could use this, but I think the hardest thing to do is really get your head around this. Here's another one. It's a fast browse. So this is really good for AI agents because usually computer use is super slow. But if we look at this example, and I think this actually came from browser use directly, what this does is it can search through flights automatically in the browser, and it's a lot cheaper and a lot faster to use. So it can quickly navigate through your browser and control it by looking at the code on the page, assign in an action to what's on the page, and then give you a quote from the page as well, which is pretty cool. It's called fast browse. So it's basically a browser agent that picks instead of generate it. So it chooses the action based on what's on the page and the code on there so that it can browse faster as an AI agent. So you will be in there with AI agents where they're like super slow to use step by step. Now, you can also use this for computer use in general, right? So there's something called Jev Desktop for Codex. And basically what this can do is allow Jev to control your computer, right? So you give it a goal, Codex plans and scopes a task, and then based on that, it can run through Jev to operate your computer way faster. And if it's something like sensitive, okay, then you have a human in the loop. But if it's something that's non-sensitive, then Jev can actually control your computer and start doing stuff for you. So for example, that could be like a specific app or a concrete goal or like text values already have. So, so many cool things you can do with Jev, lots of fun things as well. You can even use Mac OS Voice Agent with Jev.