I Challenged Jev AI to 5 Games. It Only Beat Me at .....

Author: Prompt Engineer 48 · Watch on YouTube

Video authored by the channel shown, not by IANIA; all rights belong to the creator. Transcript and summary generated automatically; the original-language content is preserved.

Original language: English

Original video: I Challenged Jev AI to 5 Games. It Only Beat Me at ….. · Channel: Prompt Engineer 48 · 17:25 · Watch on YouTube

Topics: AI agents, AI evaluation and benchmarks, AI infrastructure and hardware, AI models, Local and open-source models

Summary

The creator pits himself against Jev AI in five games: Snake, combat, runner, Flappy Bird, and Tetris. He explains how to access the service and set up a Docker container on a VPS, including obtaining an API key and connecting over SSH. In the matches, the human wins the first four games, while Jev wins Tetris by a large margin. The video concludes with a recap of the setup and links to related resources.

Key points

The five games in the challenge are Snake, combat, runner, Flappy Bird, and Tetris.
The creator describes Jev as a fast-thinking system-one model based on reinforcement learning, while noting that its training data has not been released.
He demonstrates deploying a Docker package on a Linux VPS and connecting to it through SSH.
Running the game arena requires a Jev API key, which is entered in the web interface.
The Snake match uses a modified rule that lets both players pass through their bodies and through walls.
The creator wins Snake, combat, runner, and Flappy Bird.
Jev wins Tetris, scoring 39 lines compared with the creator’s 12.
The creator says he did not run LIA on the VPS because it requires a GPU.

Chapters

Start	Chapter	Summary
00:00	The challenge and Jev overview	The creator introduces the five-game competition and briefly explains Jev and where to access it.
01:03	Setting up the VPS and Docker arena	He walks through choosing a VPS, connecting by SSH, running the Docker command, and entering an API key.
06:35	Snake and combat	The creator plays Snake and a fighting game against Jev, winning both matches.
09:42	Runner and Flappy Bird	He competes in the runner and Flappy Bird games and wins both despite describing Flappy Bird as difficult.
13:04	Tetris and final results	Jev wins Tetris, making the overall result four wins for the human and one for Jev, before the creator recaps the setup.
16:20	Resources and closing	The creator points viewers to the VPS command, API access, related videos, and explains why LIA was not run on the VPS.

Quotes

I have five games such as snake, I have a fight, I have runner, I have flappy bird and tetris. — 00:00

So, that’s a first win for me. — 08:11

I won this as well. — 13:04

I won on the four here and on the Tetris one, uh Jeff won. — 15:47

Concepts and entities

Jev AI, TypeSafe.ai, Docker, Hostinger, VPS, LIA

Full transcript (original language)

00:00 All right, in this video I have a very fun project. I’m going to compete against Jeff. So I have five games such as snake, I have a fight, I have runner, I have flappy bird and tetris. So these are the five games which I will compete. On the left will be a human which will be controlled by myself and on the right we have Jev latest. Let’s go ahead and see how we fare. In the last video, what we have done is I compared with liar which is a free alternative to dev which

00:31 is a paid solution but again it’s very cheap. Uh let’s test how it compares with human. Now if you don’t have access to dev you can go ahead to this website. It’s a typesafe.ai blog and you can go ahead. So there’s a website you can check out my other videos for explanation of this. But basically this is a system one model which means it thinks very fast. It’s not like a normal LLM where you get long responses. It’s based on reinforcement learning for calibrated decisions. What exactly that is, we don’t know because they have not

01:03 released the training data yet. But so basically, if you go to this website, go to a weight list or sign in here, then you’ll be taken here where you have home here and you have playground here. And basically, I brought you here uh because I want you to have access to the API keys that we have here. And you can see that if you go to the settings, if you go to billing, I got about $5 and I spent so much and tested so much with this. But still, I have $4.67 remaining. And it’s really great. Let’s go ahead and test this out. So for this,

01:39 what I’ve done is I have made a GitHub repo uh in the earlier video itself. But what I’ve done right now is I have made uh this package here which is like a docker package which you can use get started with the examples that you have seen here. So you can see that I am on my VPS here which is this IP. But I’m going to show you how to set it on your own system on your own IP so that you don’t have to of course you can go ahead and uh go ahead and play uh with this IP as well which is my VPS. But again you need the Jeb API key here. You just need

02:13 to put in the API key, click on save, and then you’ll have access to these uh five models uh five games that you see. But the purpose of this video is to show you how easy it is to set up on a VPS. Now, for the VPS, I’m going to use Hostinger. And you can see uh I have a link in the description. If you go to the Hostinger quickly, what Hostinger is, it provides you VPS. Uh we are going to be working with Linux. There are KVM 1, 2, 4, and 8. I’m going to choose I’m going to change the language to United States first so that you [clears throat]

02:44 get in dollars and you can see that we it’s $8.99 per month and uh the best plan would be this one KVM2 for the kind of things that we’re doing. We choose the plan and then we see that we go to this card.hostinger.com and then uh we see for 24 months here. So if we choose for 12 or 24 months then you’re going to get some discounts using the coupon code called prompt here and then clicking on apply. So you can go down and choose a server location. I’m going to choose India Mumbai because this is the uh best latency here. Now

03:19 you can go with plain OS go with Ubuntu. Confirm doesn’t matter. You just go ahead and click on continue. You can see the discounts we got here. 10% discount. Click on continue and then sign up and just make the payment. Once you have that, once you have taken a plan, what you will see is something like this. So you’ll be brought here to hpanel.hostinger.com. And if you scroll down, you will see that you have this VPS. This is my IP. This is the KVM2 plan. It’s running. This is the expiration date. You can go

03:53 ahead and click on manage to manage the session. I’m going through this quickly because I’ve done this so many times and uh I’m sure that you have listened from me on how to set this up. But you go ahead and reset this password once so that you get a fresh password. You can generate of course as well uh any complicated password that you see here. Also I want SSH uh root access. So I click here and then I go to get bash. So I’m going to paste this SSH root and you can see that I need to put in the password here. I’m going to put in the

04:26 password. You won’t see anything but this is updated now. And now we go back uh to the GitHub package that I made for you. So we want to install this. So for installing this, I have this command for you. This is just a docker uh run command. So docker rund name is arena. Then restart unless stop. Then we use this port. We’re going to use CPU 1 memory 512 MB and then uh we need to

04:57 type the uh type safe API key. So this is our API key and basically we want the API key. Now for getting the API key I’ve shown you go here uh to console.typesafe.ai and just put in like a name here and get the key. So I copy the key and I just insert it here and then we have this entire docker container entire container here which is the container that I’ve made for you. Okay. So you can go through this and uh it’s just a matter

05:28 of getting the VPS and getting and running this command here. So I’m going to take this command and I am going to go to the get bash in which we have started here. So I’m going to paste this command and I’m going to run this. Okay, you can see once I run this command, uh you will have docker running. If we do docker ps, you will see that we have this docker container which is running here. So you can see 45 minutes and we see that it’s available

06:02 in 8732 here and arena. This is the name of the container here. And now what you can do is you can go to the IP here. So this is my IP. So just go to this IP and you can go right now as well. You will get this. And then what you need to do, you need to put in the API key here. Uh as you can see I have this API key which I will remove because this is leaked now. And I’m going to paste this. Click on save and then go to this snake game. So let’s

06:35 do this. So on the left you can see that it’s human here I am I will play this game and on the right we have Jeff latest this Jeff is going to play v API these are the decisions number of decision decision per second then the latency survives here and then these are the different things which uh we have made available to Jeff because it understands because it’s based on the structured output from the list of things that it can do okay so we start

07:06 the race and let’s go. >> So you can see uh the g the the green is me and I have changed it a little bit to you know basically pass through the body as well >> and pass through walls as well. But here essentially I’m trying to get it >> as quickly as possible. Cool. You can watch. I’m already winning. You can see the scores on the

07:37 top here. It’s 60 because I’ve taken six apples and 80. And you can see the length of my snake. It’s pretty good. So, I’m essentially I’m a human. I’m smart. And actually, it looks like I’m winning this game. And I’ve changed a little bit. As I said, I can pass through the body. The same goes and it’s valid for Jeff as well. So, same rules apply. Someone commented in the last video that

08:11 the snake cannot pass through its body. But because I’m the creator here, [music] I can do whatever I want. So, score is like 140 and uh this is 90 only. So, that’s a first win for me. I go back and then I go ahead and do the combat here. [music] So, I can see on the left human. This is my health level 300%. Jab’s latest 100% health. And these are the things that it can do. Advance retreat, punch, kick, block, jump, and commit.

08:41 So, I have my AD keys to move forward. I have the F for the punch, G for the kick, H for the block, and W for the jump. So, let’s get started. I’m going to mostly jump, punch, which is W jump, uh F for punch, and G for kick. Okay, let’s do it. So, I can jump and I can kick. G for kick. Let’s kick him. Yep.

09:12 Kick. I’m not even letting it to hit me even once. I’m going to punch now. Okay. Punch. Punch. Nice. Looks like I’m losing. No, [music] I’m winning. Cool. I win. So, two of the games I win. I go back and [music] then this is the

09:42 runner. This is a little difficult for me. Let’s go back and uh go ahead. And there are three lives here. Let’s run. So, I can do a N and D. So, on the left, it’s me. On the right is Jeff latest and I can see that we both are pretty much pretty good. Okay, Jev lost one life there, but I’m still here standing strong.

10:13 >> Good. The speed is going to increase once we you know it’s going to increase once we complete a certain distance and then it will be further difficult to control it faster. Still it’s me winning here. So two life lost and I win this

10:44 right. So in this as well uh the human is the one which uh who is winning and you can see the speed is increasing but I’m going to play it through until and unless [music] I lost I lose my three lives here. Good good. Can you see my speed? All real time, baby.

11:24 Oh my god. One life lost. Nice, nice, nice, nice life lost. One remaining. >> [music] >> gone. So, distance is like 3 kilometers, 3,000. It’s like 800 m. Come on. I want this. Let’s go back to the Flappy Bird. This is one of the tough games. Uh, I

11:57 would say again, three lives. And I don’t like this game. Okay, let’s go ahead and try this out. Uh, space is for jumping and I can say fly. I hate this game because it’s really difficult for me. So on the left gone, one life gone, but I really see very good potential. Uh on the right, you can see Jeff is playing really well this game.

12:27 Nice. Let’s see. Let’s see. Come on. I hate this game, man. The acceleration and de acceleration is like too fast. The gravity is gone. The gravity is really strong here. And this bird doesn’t have gliding power. It’s like a small bird. Goes up and goes down pretty fast. It’s not like an eagle uh which can soar.

13:04 Okay. One down. two. I think I’m going to win again. Don’t know. Let’s see. It’s a never- ending game. Okay, come on. I won this as well. So, let’s go back and uh let’s uh do this Tetris here. So again on the left it’s me on the right is Jeff and I’m going to play

13:35 this. Let’s do this. This little this is a little difficult and you can see that it’s pretty fast here. So I get like a few seconds to decide and it comes down immediately or you can pull it down as well. But certainly you can see that Jev is winning here. But let’s see till the end.

14:09 Okay, let’s see who wins. I’m really liking this. I have been a player as well. Come on, man. [music] Waiting for a straight line. Come on. Okay.

14:42 Okay. Okay. I can put it here. Sorry. Just missed it. Okay. Done. Oh my god, that’s a miss. Oh my god, I’m losing my confidence. Yep. Still, let’s see who wins. [music]

15:14 It’s the one. Okay. All right. All right. All right. All right. All right. I lost. I conquer. I lost because I have only 12 lines

15:47 here. And on the right, you can see it’s 39 lines. Really amazing. Very powerful. Very good. Cool. So that’s the one five games. Uh I won on the four here and on the Tetris one, uh Jeff won. Basically I lost. So sad. But in summary, what we have done is we went on to take a KVM2 plan and then we went ahead and SSH into that. And after

16:20 that we ran this beautiful command or docker container here because I’ve already made the docker container for you. This is the container as you can see we just run the command and we have the docker running. So once we have the docker running it’s on port 80 here and therefore it will open u itself here if you just put the IP. Now I can share this with you as well but I want you to go ahead and create on your own system. uh take a VPS plan and then run it there. Also, if you want to use my system, you need to put in the API key

16:52 here. You can go ahead and get the API key here. Also, if you want to learn more about uh using VPS. So, I have this uh good video of build your own uh cloud 10 projects one VPS. You can go ahead and check this out as well. You can check out the intro video to dev uh just published one day ago. And this liar which is a free version versus uh Jeff paid version comparison as well. Now important thing why I didn’t choose uh liar to be run on the VPS is because we

17:25 need a GPU to run LIA which can run on my own system because I have an 8GB GPU uh which I’ve shown here. You can go ahead and test this out and again the codes are available and all the details are available. You can go ahead and check this out. Subscribe to my channel. I publish videos almost every day because so many different things that we have on the internet, we have on AI and I’m really excited to bring this things to you. So, go ahead and check out and I’ll see you in the next