1 00:00:00,020 --> 00:00:03,700 Josh: A few months ago, we told you to use Claude. Now we're telling you to switch 2 00:00:03,700 --> 00:00:06,720 Josh: back because for those of you who aren't familiar, well, over Christmas break, 3 00:00:06,820 --> 00:00:09,820 Josh: there was a major vibe shift where AI coding went from this like fun tool to 4 00:00:09,820 --> 00:00:12,400 Josh: things that developers actually use when they're shipping code. 5 00:00:12,600 --> 00:00:15,460 Josh: And even if you're not a developer, the amount of use cases and applications 6 00:00:15,460 --> 00:00:18,000 Josh: that were created around that time were really strong. 7 00:00:18,200 --> 00:00:21,360 Josh: And since then, Anthropic has gone on this generational run of shipping these 8 00:00:21,360 --> 00:00:24,920 Josh: incredible products seemingly every single day that has turned Claude code into 9 00:00:24,920 --> 00:00:28,480 Josh: this supercharged super app that is the place that EJS, 10 00:00:28,540 --> 00:00:33,220 Josh: I know you've gone to, I've gone there too, in order to get all of our AI progress done. 11 00:00:33,360 --> 00:00:35,340 Josh: Any work that we have, we've gone to Cloud Code. 12 00:00:35,500 --> 00:00:39,000 Josh: Now, OpenAI has woken up. And over the last few weeks, Codex has shipped more 13 00:00:39,000 --> 00:00:41,440 Josh: features than most companies ship in a year. 14 00:00:41,500 --> 00:00:44,480 Josh: And I bet, I guarantee that you haven't heard of some of these features that 15 00:00:44,480 --> 00:00:45,860 Josh: we're going to talk about in this episode. 16 00:00:46,060 --> 00:00:49,820 Josh: The pendulum has fully swung back, or at least I believe, because I'm totally Codex-pilled. 17 00:00:49,980 --> 00:00:53,200 Josh: And in this episode, we're going to kind of walk through the differences between 18 00:00:53,200 --> 00:00:57,760 Josh: these two and why the model that you're using today probably won't be the model you're using tomorrow. 19 00:00:58,000 --> 00:01:00,840 Josh: And I don't think we're going to convince you, but maybe we could show you why 20 00:01:00,840 --> 00:01:03,740 Josh: you might want to consider using something else here. 21 00:01:03,920 --> 00:01:06,880 Ejaaz: I just want to talk through some of the crazy stats here because the script 22 00:01:06,880 --> 00:01:11,660 Ejaaz: has genuinely flipped. A few months ago, Claude Code was anything everyone could talk about. 23 00:01:11,820 --> 00:01:17,520 Ejaaz: And every software engineer was using Claude Code. Every enterprise was installing it. It was crazy. 24 00:01:17,660 --> 00:01:21,200 Ejaaz: But just over the last couple of weeks, specifically by the end of April, 25 00:01:21,730 --> 00:01:26,250 Ejaaz: Chat GPT 5.5 was released, and that was plugged into the coding AI model. 26 00:01:26,390 --> 00:01:27,570 Ejaaz: It's all one and the same. 27 00:01:27,950 --> 00:01:31,190 Ejaaz: And OpenAI went on this code red run where they focus on nothing but building 28 00:01:31,190 --> 00:01:33,990 Ejaaz: the best coding AI model and the best LLM. 29 00:01:34,150 --> 00:01:36,030 Ejaaz: And the numbers show that it's worked. 30 00:01:36,210 --> 00:01:41,950 Ejaaz: Over the last week, Codex has been downloaded over or installed over 46 million times. 31 00:01:42,290 --> 00:01:46,890 Ejaaz: Cold code, under 500,000 times. Now, that is crazy to say, because if you look 32 00:01:46,890 --> 00:01:51,310 Ejaaz: at the historical data, cold code downloads and installs has absolutely dwarfed 33 00:01:51,310 --> 00:01:53,810 Ejaaz: Codex, but something changed over the last couple of weeks. 34 00:01:53,930 --> 00:01:58,930 Ejaaz: That something was OpenAI putting out just a better model. You mentioned that 35 00:01:58,930 --> 00:02:00,030 Ejaaz: you were Codex pill, Josh. 36 00:02:00,190 --> 00:02:04,750 Ejaaz: I think so am I. I've spent the last couple of days playing around with Codex. 37 00:02:04,830 --> 00:02:09,350 Ejaaz: This morning, we prepped a bunch of really cool demos, and it is just completely flipped script. 38 00:02:09,530 --> 00:02:13,090 Ejaaz: But it's one thing saying it. It's another thing actually showing the direct 39 00:02:13,090 --> 00:02:15,990 Ejaaz: comparison. So we created this visual artifact to kind of 40 00:02:16,370 --> 00:02:18,890 Ejaaz: give you the scoreboard. And you can see it at the top here. 41 00:02:19,070 --> 00:02:23,710 Ejaaz: It's OpenAI Codex at 11 and Anthropic Claude at two. But let me explain why. 42 00:02:23,910 --> 00:02:26,290 Ejaaz: Okay. So number one, computer use. 43 00:02:26,830 --> 00:02:30,370 Ejaaz: Codex and Claude code can use your computer. 44 00:02:30,570 --> 00:02:33,290 Ejaaz: It can take over your desktop and it can like move your cursor around. 45 00:02:33,410 --> 00:02:37,950 Ejaaz: Now, Claude pioneered it. There were the first ones there, but it was super slow. 46 00:02:38,170 --> 00:02:41,610 Ejaaz: It kind of runs into a bunch of obstacles and you have to kind of like handhold 47 00:02:41,610 --> 00:02:43,510 Ejaaz: it and improve it to do a bunch of different things. 48 00:02:43,950 --> 00:02:47,770 Ejaaz: Codex is not only quicker than me, it's quicker than the average person. 49 00:02:47,910 --> 00:02:50,670 Ejaaz: In fact, I can actually see the cursor move around so quickly. 50 00:02:50,710 --> 00:02:55,890 Ejaaz: And it's like using a computer, but it's a superhuman and it can run pretty much 24-7 at this point. 51 00:02:56,350 --> 00:03:01,570 Ejaaz: Long horizon autonomy. Codex can work for longer in a much more intelligent 52 00:03:01,570 --> 00:03:05,750 Ejaaz: manner versus Claude code, which is, again, crazy to say because literally a 53 00:03:05,750 --> 00:03:07,790 Ejaaz: month ago, it was the inverse of this. 54 00:03:07,950 --> 00:03:11,630 Ejaaz: Claude right now can run for a decent number of times or amount of time, 55 00:03:11,730 --> 00:03:13,810 Ejaaz: but not as long as Codex can. 56 00:03:14,250 --> 00:03:17,090 Ejaaz: And then the last two that I want to talk about here is browser use. 57 00:03:17,290 --> 00:03:22,010 Ejaaz: So Codex can take over your browser. It can do a lot more intentional things. 58 00:03:22,070 --> 00:03:25,670 Ejaaz: It understands what it's looking at, very importantly. Previously, it could not do that. 59 00:03:25,970 --> 00:03:29,130 Ejaaz: Claude can do the same, but not as intelligently. And then finally, 60 00:03:29,670 --> 00:03:35,510 Ejaaz: ChatGPT Images 2.0 got released, what was it, like two weeks ago now? Oh, it's so good. 61 00:03:35,610 --> 00:03:40,070 Ejaaz: Yeah, it's the image generation model from OpenAI, and it is absolutely astounding. 62 00:03:40,170 --> 00:03:44,330 Ejaaz: In fact, it beat all the other predecessors, including Google's, 63 00:03:44,330 --> 00:03:49,970 Ejaaz: what is it, Nano Banana 2.0 Pro, which previously held the lead. 64 00:03:50,110 --> 00:03:51,890 Ejaaz: It beat it across every single benchmark. 65 00:03:52,370 --> 00:03:54,970 Ejaaz: Anthropic, on the other hand, doesn't even have an image gen model. 66 00:03:55,110 --> 00:03:57,270 Ejaaz: So, so far, it's crushing. Yeah. 67 00:03:57,650 --> 00:04:01,030 Josh: Yeah, I think a lot of the best is now bundled into codex. The image gen for 68 00:04:01,030 --> 00:04:04,590 Josh: anyone who uses any sort of visual work is unbelievable. And being able to use 69 00:04:04,590 --> 00:04:07,050 Josh: that directly in your software is awesome. 70 00:04:07,390 --> 00:04:10,630 Josh: One thing that you mentioned is the long horizon autonomy. I think that needs 71 00:04:10,630 --> 00:04:14,410 Josh: a double clicking on because it's really impressive how well it works. 72 00:04:14,830 --> 00:04:17,090 Josh: Traditionally, there's been this thing called a Ralph loop that we use. 73 00:04:17,190 --> 00:04:20,230 Josh: It's actually named after the character from The Simpsons who is very persistent. 74 00:04:20,470 --> 00:04:25,130 Josh: And it's basically a planning mode where you give the AI a goal and it will 75 00:04:25,130 --> 00:04:27,550 Josh: continue to iterate towards that goal until it accomplishes it. 76 00:04:27,630 --> 00:04:32,030 Josh: So like, let's say you want to build a Lego car or something and you give it the exact parameters. 77 00:04:32,130 --> 00:04:35,150 Josh: It will go and go and go until it solves that problem and gives you exactly 78 00:04:35,150 --> 00:04:37,470 Josh: what you want in a way that other AI models haven't. 79 00:04:37,730 --> 00:04:41,090 Josh: Codex did that. And this is the only native implementation that you can get 80 00:04:41,090 --> 00:04:44,950 Josh: of this long horizon thinking where it actually will go for days on end. 81 00:04:45,030 --> 00:04:48,610 Josh: I've seen screenshots of some thinking for as long as 36 hours to accomplish the goal. 82 00:04:48,710 --> 00:04:52,590 Josh: So if you have really difficult tasks, Codex is going to be really good at solving those. 83 00:04:52,670 --> 00:04:55,390 Josh: Now, continuing to scroll down, there was another feature that was just released 84 00:04:55,390 --> 00:04:56,950 Josh: this week called auto review. 85 00:04:57,170 --> 00:05:01,190 Josh: And a huge pain in the ass for people who are creating code for working on complex 86 00:05:01,190 --> 00:05:05,170 Josh: projects, whatever it may be, is you're constantly having to sit there and approve 87 00:05:05,170 --> 00:05:07,750 Josh: things because the permission system is a little finicky, right? 88 00:05:07,810 --> 00:05:09,370 Josh: You don't want to give it full access to your computer. 89 00:05:10,180 --> 00:05:13,160 Josh: You also don't want to sit there and approving every time it wants to use Chrome 90 00:05:13,160 --> 00:05:14,740 Josh: or every time I want to access your file. 91 00:05:15,200 --> 00:05:19,940 Josh: So Codex created auto review and they rolled it out last week where the agent is kind of smart. 92 00:05:20,100 --> 00:05:24,980 Josh: It knows which things are going to possibly be systemic existential threats 93 00:05:24,980 --> 00:05:26,060 Josh: and which approvals aren't. 94 00:05:26,140 --> 00:05:29,900 Josh: And it will just automatically approve all the things that aren't going to get you in a lot of trouble. 95 00:05:30,020 --> 00:05:33,120 Josh: It creates a much easier user interface where you can just kind of walk away 96 00:05:33,120 --> 00:05:35,940 Josh: from the computer for a little while and come back and things get done. 97 00:05:35,940 --> 00:05:38,680 Josh: Memory and context is pretty strong i'd say 98 00:05:38,680 --> 00:05:41,520 Josh: the one thing and we haven't mentioned many claude winners the 99 00:05:41,520 --> 00:05:44,780 Josh: place where claude wins currently is on their open claw capability funny 100 00:05:44,780 --> 00:05:49,060 Josh: enough because open ai bought open claw but dispatch is the mobile app feature 101 00:05:49,060 --> 00:05:53,480 Josh: for claude in which you can actually engage with claude code remotely that doesn't 102 00:05:53,480 --> 00:05:57,020 Josh: currently exist on codex and while the team has promised to ship that you don't 103 00:05:57,020 --> 00:06:01,160 Josh: actually have that currently today claude has that also in terms of the personality 104 00:06:01,160 --> 00:06:04,440 Josh: and ui claude is just so much better i think we're going to get into our. 105 00:06:04,720 --> 00:06:08,180 Josh: Personal takes but whenever you're using an llm versus an actual 106 00:06:08,180 --> 00:06:10,980 Josh: tool set or a harness claude is pretty great and the 107 00:06:10,980 --> 00:06:13,740 Josh: ui is very warm so there's there's some kind of 108 00:06:13,740 --> 00:06:17,140 Josh: instances in which claude is better but for the most part codex is 109 00:06:17,140 --> 00:06:20,220 Josh: really just kind of crushing it and i've really enjoyed using it one of the 110 00:06:20,220 --> 00:06:25,400 Josh: fun things is pets i mean just recently they released pets and claude also released 111 00:06:25,400 --> 00:06:28,800 Josh: pets but these pets are a little bit different this is an example of angry dario 112 00:06:28,800 --> 00:06:32,260 Josh: we're seeing on the screen and it's fun because you have this persistent character 113 00:06:32,260 --> 00:06:34,740 Josh: that exists throughout your computer use. 114 00:06:34,940 --> 00:06:38,440 Josh: And as you're engaging with Codex, it'll just kind of chat with you in the background 115 00:06:38,440 --> 00:06:39,920 Josh: so you can see your progress, see where you're at. 116 00:06:40,040 --> 00:06:43,220 Josh: It's fun, it's playful, and it just shows that they kind of care about the user experience. 117 00:06:43,380 --> 00:06:47,240 Josh: Now, one feature I would guarantee most people don't know is Chronicle, EJS. 118 00:06:47,300 --> 00:06:48,920 Josh: And you were just telling me about Chronicle and how cool it is, 119 00:06:49,000 --> 00:06:52,080 Josh: how it kind of monitors your screen as you go. This seems like novel technology 120 00:06:52,080 --> 00:06:53,100 Josh: that we haven't seen yet. 121 00:06:53,380 --> 00:06:58,340 Ejaaz: Yeah, so one of the earliest episodes that we did here on Limitless was an interview 122 00:06:58,340 --> 00:07:00,740 Ejaaz: with the folks at OpenAI that created... 123 00:07:01,240 --> 00:07:03,880 Ejaaz: Something called, what was it called, Josh? Do you remember? 124 00:07:04,080 --> 00:07:07,720 Ejaaz: It was like agent mode or personal mode, something like that. 125 00:07:08,240 --> 00:07:10,060 Josh: Yes. It thought overnight for you, right? Yes. 126 00:07:10,200 --> 00:07:15,240 Ejaaz: It basically took all the conversations that you'd had with ChatGPT the night 127 00:07:15,240 --> 00:07:20,020 Ejaaz: before or the day before or the week before, and it created important context 128 00:07:20,020 --> 00:07:22,860 Ejaaz: around you in the form of something called memories. 129 00:07:22,860 --> 00:07:27,480 Ejaaz: This is where AI memory was birthed from OpenAI themselves, from the OpenAI 130 00:07:27,480 --> 00:07:31,720 Ejaaz: team. And what it would do is it would feed you a report in the morning that 131 00:07:31,720 --> 00:07:35,340 Ejaaz: would update you on information that it thought you would be interested to read about. 132 00:07:35,520 --> 00:07:37,900 Ejaaz: So say, for example, you were interested in the stock market, 133 00:07:37,920 --> 00:07:40,900 Ejaaz: it'll give you an update on a bunch of advancements that had happened overnight 134 00:07:40,900 --> 00:07:42,820 Ejaaz: or over the last week or whatever it might be. 135 00:07:43,140 --> 00:07:47,580 Ejaaz: Right now, fast forward today, memory is embedded across every single AI model 136 00:07:47,580 --> 00:07:50,780 Ejaaz: and tool. The reason why is context is so important. 137 00:07:50,940 --> 00:07:54,300 Ejaaz: It's one thing a user asking for something explicitly and directly. 138 00:07:54,300 --> 00:07:58,160 Ejaaz: It's a complete other thing for an AI to actually understand what you mean, 139 00:07:58,400 --> 00:08:01,420 Ejaaz: the nuance in the sentence that you've created, and even better, 140 00:08:01,660 --> 00:08:03,320 Ejaaz: to predict what you want. 141 00:08:03,460 --> 00:08:07,760 Ejaaz: But there was still an obstacle, which was you needed to feed it the context 142 00:08:07,760 --> 00:08:10,780 Ejaaz: and say, hey, Claude, hey, ChatGPT, can you remember this? 143 00:08:11,060 --> 00:08:18,340 Ejaaz: OpenAI recently released a feature called Chronicle, where it observes what you scroll through, 144 00:08:18,700 --> 00:08:23,740 Ejaaz: What you click on, what you type, and it builds its own context and memories 145 00:08:23,740 --> 00:08:28,120 Ejaaz: around you without you needing to feed it, which actually led to a really cool 146 00:08:28,120 --> 00:08:31,340 Ejaaz: prompt that you pointed out, Josh, or that you found, which was, 147 00:08:31,570 --> 00:08:35,170 Ejaaz: what have I been doing very inefficiently on my computer, according to Chronicle, 148 00:08:35,510 --> 00:08:38,230 Ejaaz: which is this new memory feature, make some recommendations, 149 00:08:38,490 --> 00:08:41,890 Ejaaz: be direct, tell me what I need to hear. That's, that's pretty awesome. 150 00:08:42,170 --> 00:08:45,270 Josh: Yeah. So this is alpha because I don't think a lot of people recognize that 151 00:08:45,270 --> 00:08:49,010 Josh: this is a possibility because Codex and OpenAI didn't do a good job of explaining this. 152 00:08:49,110 --> 00:08:52,550 Josh: When they released Chronicle, they said it's a way of the system to review your 153 00:08:52,550 --> 00:08:55,830 Josh: code as you've gone because it's been taking sequential screenshots. 154 00:08:55,990 --> 00:08:58,130 Josh: But it's the reality is, is that it's much bigger than this. 155 00:08:58,170 --> 00:09:01,710 Josh: And I suspect they didn't market it this way because it could be a bit of a privacy issue, 156 00:09:01,710 --> 00:09:05,210 Josh: but it's essentially constantly monitoring your screen and taking screenshots 157 00:09:05,210 --> 00:09:09,130 Josh: of what's happening on your screen and interpreting it so it understands your 158 00:09:09,130 --> 00:09:11,530 Josh: habits, the way that you work, the thing that you do. 159 00:09:11,750 --> 00:09:14,750 Josh: And then you can ask it, what have I been doing very inefficiently on my computer? 160 00:09:14,890 --> 00:09:17,190 Josh: According to Chronicle, make some recommendations, be direct, 161 00:09:17,350 --> 00:09:18,150 Josh: tell me what I need to hear. 162 00:09:18,250 --> 00:09:21,110 Josh: And it'll actually evaluate how you've been using your computer, 163 00:09:21,310 --> 00:09:24,570 Josh: how long you've been scrolling on Twitter, perhaps, how long you haven't been 164 00:09:24,570 --> 00:09:27,130 Josh: doing the things you're supposed to be working on, or just generally how to 165 00:09:27,130 --> 00:09:31,890 Josh: improve your workflow and give you real feedback based on your actual actions that it's seen. 166 00:09:31,990 --> 00:09:35,370 Josh: And I think this is a super powerful thing currently only available to pro members. 167 00:09:35,450 --> 00:09:39,490 Josh: So if you pay for the $100, $200 a month subscription, you get access to this. 168 00:09:39,590 --> 00:09:42,970 Josh: But I suspect this is the early signs of a very important feature they're going 169 00:09:42,970 --> 00:09:48,470 Josh: to roll out, which is that entire computer monitoring system to improve your 170 00:09:48,470 --> 00:09:51,650 Josh: system and also probably train the models to get better at engaging with your system. 171 00:09:51,750 --> 00:09:54,910 Josh: But I found Chronicle to be one of those kind of secret features that not a 172 00:09:54,910 --> 00:09:57,830 Josh: lot of people know about, but has a lot of upside if you use it to your advantage 173 00:09:57,830 --> 00:10:01,070 Josh: and let it monitor what you're doing and improve your workflow on a day-to-day basis. 174 00:10:01,650 --> 00:10:05,110 Ejaaz: So the point is, from both of these companies, Anthropic and OpenAI, 175 00:10:05,290 --> 00:10:09,470 Ejaaz: we are getting feature releases every single week. In fact, every single day. 176 00:10:09,910 --> 00:10:13,010 Ejaaz: And it's becoming, I'm being bombarded by this. 177 00:10:13,150 --> 00:10:16,630 Ejaaz: And it's hard to keep track with all of this. So what is the number one litmus 178 00:10:16,630 --> 00:10:19,350 Ejaaz: test for both of these models and products and companies? 179 00:10:19,670 --> 00:10:23,090 Ejaaz: It's to actually use the thing. It's to build the thing. 180 00:10:23,330 --> 00:10:27,470 Ejaaz: And we have two special demos that we have prepared for you that we're about 181 00:10:27,470 --> 00:10:30,550 Ejaaz: to jump into. Now, Josh, can you guess what my first demo is about? 182 00:10:30,690 --> 00:10:33,430 Josh: The theme. First one's a game. We're gamers, man. I want to play a game. 183 00:10:33,570 --> 00:10:34,810 Josh: I want to see how well it does on a game. 184 00:10:34,970 --> 00:10:38,630 Josh: I know we did this demo in the past months ago. It left a lot to be desired. 185 00:10:38,710 --> 00:10:42,550 Josh: So I'm curious to see the current up-to-date status as it relates to Cloud Code 186 00:10:42,550 --> 00:10:45,410 Josh: versus Codex. Who's winning on the one-shot game prompt? 187 00:10:45,750 --> 00:10:49,690 Ejaaz: Indeed. Okay. So I am a nostalgic kind of guy. And so I was like, 188 00:10:49,810 --> 00:10:51,830 Ejaaz: oh, back in the day, I loved Mario. 189 00:10:52,070 --> 00:10:58,110 Ejaaz: So I want you, both of these models, to create the best Mario type or inspired 190 00:10:58,110 --> 00:11:00,110 Ejaaz: game, a side scroller, but make it futuristic. 191 00:11:00,780 --> 00:11:03,560 Ejaaz: Maybe add a little bit of neon, sprinkle a bit of neon in there, 192 00:11:04,020 --> 00:11:06,960 Ejaaz: create levels. I want game design. I want there to be enemies. 193 00:11:07,080 --> 00:11:08,200 Ejaaz: I want there to be pitfalls. 194 00:11:08,380 --> 00:11:11,560 Ejaaz: And I also want there to be a scoreboard and also tell me how to do this thing. 195 00:11:11,640 --> 00:11:13,020 Ejaaz: I want, give me the whole package. 196 00:11:13,140 --> 00:11:17,940 Ejaaz: Basically, I fed this prompt or idea into ChatGPT and Claude. 197 00:11:18,020 --> 00:11:21,740 Ejaaz: And I said, can you create a detailed prompt that I can then feed into your coding models? 198 00:11:21,920 --> 00:11:25,300 Ejaaz: I then set each of the coding models to their highest settings. 199 00:11:25,480 --> 00:11:28,200 Ejaaz: So what you're about to see is the best of the best for the most 200 00:11:28,200 --> 00:11:30,960 Ejaaz: detailed prompt that they came up with and let's see what they 201 00:11:30,960 --> 00:11:33,960 Ejaaz: did so step number one or example number 202 00:11:33,960 --> 00:11:37,260 Ejaaz: one is called opus 4.7 so 203 00:11:37,260 --> 00:11:40,440 Ejaaz: this is called code at the highest setting with their latest model 204 00:11:40,440 --> 00:11:46,240 Ejaaz: um okay it took the prompt pretty literally it's titled this neon plumber moon 205 00:11:46,240 --> 00:11:49,940 Ejaaz: base run which is obviously mario inspired and it said hey this is a demo edition 206 00:11:49,940 --> 00:11:53,160 Ejaaz: by the way this is not production ready what i like about this is it's giving 207 00:11:53,160 --> 00:11:57,500 Ejaaz: me the instructions but how does the game actually play out let's see it looks 208 00:11:57,500 --> 00:11:59,420 Ejaaz: good can you see me here josh i 209 00:11:59,420 --> 00:12:01,580 Josh: Can yes i can and it looks like. 210 00:12:01,580 --> 00:12:04,300 Ejaaz: The animations are pretty good i'm jumping around i think 211 00:12:04,300 --> 00:12:07,380 Ejaaz: i'm like a little robot i can see my feet pitter-pattering now 212 00:12:07,380 --> 00:12:10,980 Ejaaz: i'm guessing this thing is about to kill me so let's see if i can jump oh i 213 00:12:10,980 --> 00:12:16,420 Ejaaz: can jump there we go that's awesome um one bit can i kill this guy oh yes i 214 00:12:16,420 --> 00:12:22,180 Ejaaz: can now one bit of feedback i've noticed is uh i can't double jump and it told 215 00:12:22,180 --> 00:12:23,460 Ejaaz: me in the menu that i could double 216 00:12:23,460 --> 00:12:26,800 Ejaaz: jump so that's weird so the physics hasn't really paid off can i die 217 00:12:28,650 --> 00:12:30,590 Josh: Oh, it certainly looks like you could die. 218 00:12:30,770 --> 00:12:35,650 Ejaaz: I can die. Great. Okay. So that is Claude's attempt at it. What's your feedback 219 00:12:35,650 --> 00:12:37,350 Ejaaz: on this, Josh? I think the graphics are pretty good. 220 00:12:37,470 --> 00:12:41,090 Josh: The graphics are great. For one shot, I mean, granted, this is only one single 221 00:12:41,090 --> 00:12:43,310 Josh: prompt. So for one prompt, it created great graphics. 222 00:12:43,470 --> 00:12:46,670 Josh: It had sound design that actually sounds pretty accurate to what you would expect in the game. 223 00:12:46,810 --> 00:12:49,990 Josh: It has similar principles. It's following gaming principles. 224 00:12:50,130 --> 00:12:52,710 Josh: You kind of understand what looks dangerous, what doesn't. 225 00:12:52,910 --> 00:12:55,130 Josh: You knew that those spikes were going to hurt you and they hurt you. 226 00:12:55,130 --> 00:13:00,210 Josh: The logic seems to be a little bit flawed i think it's having problems with gravity or at least that 227 00:13:00,430 --> 00:13:03,310 Josh: double jump functionality because it looks like those coins that you probably 228 00:13:03,310 --> 00:13:06,390 Josh: want to collect you can't actually reach because you can't do the double jump 229 00:13:06,390 --> 00:13:11,190 Josh: so in terms of logic not so hot in terms of visuals aesthetics in terms of i 230 00:13:11,190 --> 00:13:14,590 Josh: mean how good this game is from one shot very impressive yeah. 231 00:13:14,590 --> 00:13:20,030 Ejaaz: I think it's important to understand that i started from zero it literally asked 232 00:13:20,030 --> 00:13:23,430 Ejaaz: me to give it a folder to build in and the folder was completely empty. 233 00:13:23,670 --> 00:13:27,470 Ejaaz: So all the visual renderings, all the graphics, the animation style, 234 00:13:27,630 --> 00:13:32,410 Ejaaz: the scoring system, the way that the avatar moves and looks was created from 235 00:13:32,410 --> 00:13:34,890 Ejaaz: scratch from a bunch of characters from this AI model. 236 00:13:35,010 --> 00:13:39,110 Ejaaz: So this is Claude Code's current best attempt and it is way better than what 237 00:13:39,110 --> 00:13:42,870 Ejaaz: we tested out and honestly demoed on this show about a month ago. 238 00:13:42,970 --> 00:13:49,430 Ejaaz: But now let's see what OpenAI's ChatGPT 5.5 codex at the highest possible setting cooked up 239 00:13:49,430 --> 00:13:52,350 Josh: Okay and this is using the same prompt correct so you just fed the model the 240 00:13:52,350 --> 00:13:56,510 Josh: same prompt identical but identical right oh god i'm excited i hope codex did 241 00:13:56,510 --> 00:13:59,930 Josh: well because now now that i'm a fan i'm gassing it up it better perform here okay. 242 00:13:59,930 --> 00:14:02,630 Ejaaz: So this is gpt 5.5's attempt now you 243 00:14:02,630 --> 00:14:05,610 Ejaaz: might notice that this isn't the entire browser that's because 244 00:14:05,610 --> 00:14:08,770 Ejaaz: codex has a very unique feature which is not only 245 00:14:08,770 --> 00:14:11,390 Ejaaz: can it do all the coding in a single app for you but it 246 00:14:11,390 --> 00:14:14,110 Ejaaz: has an in-app browser so it can 247 00:14:14,110 --> 00:14:17,770 Ejaaz: live test the thing in the app without you needing to go to google chrome or 248 00:14:17,770 --> 00:14:22,710 Ejaaz: whatever but anyway we have the starting screen here it has also called it neo 249 00:14:22,710 --> 00:14:27,210 Ejaaz: neon plumber moon base run it looks a little more rudimentary from the start 250 00:14:27,210 --> 00:14:31,110 Ejaaz: but i do like the background animation josh we didn't get this in the previous 251 00:14:31,110 --> 00:14:33,150 Ejaaz: one or at least not this side scrolling thing well let's 252 00:14:33,740 --> 00:14:34,120 Ejaaz: Oh. 253 00:14:34,500 --> 00:14:35,480 Josh: Oh, this is nice. 254 00:14:35,740 --> 00:14:38,680 Ejaaz: This is nice. I think this has good logic. 255 00:14:39,360 --> 00:14:43,640 Ejaaz: Wait, but this is no music. There's no music. I can't double jump. 256 00:14:44,100 --> 00:14:46,320 Ejaaz: Might be a skill issue. Might be a prompt issue. 257 00:14:46,580 --> 00:14:48,620 Josh: Let's have a look. Did it say you can double jump? 258 00:14:49,140 --> 00:14:50,380 Ejaaz: That's a good question, actually. 259 00:14:50,500 --> 00:14:52,940 Josh: This is a fully playable game. 260 00:14:53,120 --> 00:14:57,960 Ejaaz: Yes. And I like that it's like zoomed in. There's like... Oh, 261 00:14:58,040 --> 00:15:00,820 Ejaaz: we got the boost. I can jump on the platforms. Let's see if I can kill this guy. 262 00:15:02,520 --> 00:15:03,720 Josh: Yes. nice okay. 263 00:15:03,720 --> 00:15:05,720 Ejaaz: And can i jump the gap there's a scoring system 264 00:15:05,720 --> 00:15:09,040 Josh: You could see your hearts oh dude this is way better power. 265 00:15:09,040 --> 00:15:13,300 Ejaaz: Up wait oh my god i want the power up i'm still gonna go back double jump 266 00:15:13,300 --> 00:15:16,540 Josh: You can you could go back go back to the last platform oh god. 267 00:15:16,540 --> 00:15:19,000 Ejaaz: I died i'm going i'm going to the last platform here 268 00:15:19,000 --> 00:15:22,380 Josh: We go it looks like they're sequentially gaining height which is interesting, 269 00:15:23,000 --> 00:15:25,920 Josh: oh but okay so if i'm comparing these two i'm actually i'm not feeling very 270 00:15:25,920 --> 00:15:29,320 Josh: let down this is good aside from the music not existing which we may not have 271 00:15:29,320 --> 00:15:33,520 Josh: explicitly asked um this it looks like the logic plays better the actual gameplay 272 00:15:33,520 --> 00:15:38,660 Josh: is usable this is a full i don't know if it's glitching or if this is you glitching no no. 273 00:15:38,660 --> 00:15:40,460 Ejaaz: That is it's glitching it's glitching a bit 274 00:15:40,460 --> 00:15:43,300 Josh: Okay so it's still there are some edge case errors yeah but 275 00:15:43,300 --> 00:15:46,600 Josh: this is different in the sense that you have your hearts clearly projected you 276 00:15:46,600 --> 00:15:49,920 Josh: have a score system that's clearly in place you're able to get these power-ups 277 00:15:49,920 --> 00:15:53,620 Josh: they work they function i mean this is a very clean and functional game so i 278 00:15:53,620 --> 00:15:58,880 Josh: would give this to codex i think the experience perhaps the design of claude 279 00:15:58,880 --> 00:16:01,440 Josh: was better and And perhaps the music, 280 00:16:01,600 --> 00:16:03,820 Josh: I mean, music was definitely better versus none, but Claude, 281 00:16:04,000 --> 00:16:07,000 Josh: in terms of just, or Codex, in terms of just coding logic and making a better 282 00:16:07,000 --> 00:16:09,860 Josh: game, I give, I give this Codex. Do you have a take? 283 00:16:10,580 --> 00:16:15,600 Ejaaz: Yeah. So on the build side of things, I had a much more pleasant experience. 284 00:16:16,470 --> 00:16:19,230 Ejaaz: Using codex as well so i think codex wins 285 00:16:19,230 --> 00:16:22,230 Ejaaz: on this um i one-shotted it in the true sense 286 00:16:22,230 --> 00:16:24,930 Ejaaz: where i just gave it a single prompt and codex didn't ask for 287 00:16:24,930 --> 00:16:28,410 Ejaaz: any permissions it just kind of went on and did the thing i saw 288 00:16:28,410 --> 00:16:34,930 Ejaaz: it it's thinking and at points where it was unsure it thought amongst itself 289 00:16:34,930 --> 00:16:39,110 Ejaaz: and then made the decision to progress forwards whereas with cloud code it would 290 00:16:39,110 --> 00:16:43,610 Ejaaz: come to me now that might just be a developer engineer's preference right like 291 00:16:43,610 --> 00:16:45,770 Ejaaz: if you're building a production ready app for like, I don't know, 292 00:16:46,050 --> 00:16:49,890 Ejaaz: a big company that you work for, you probably want to have more hands-on involvement. 293 00:16:50,070 --> 00:16:52,670 Ejaaz: Whereas if you're just building a game like we did today, where I don't really 294 00:16:52,670 --> 00:16:56,710 Ejaaz: care what it ends up looking like or what it does, then the hands-off preference 295 00:16:56,710 --> 00:17:00,770 Ejaaz: is probably something that you would use Codex for. But I think Codex wins this. 296 00:17:01,290 --> 00:17:04,890 Josh: So for our second demo, we have this handwritten piece of paper that I actually 297 00:17:04,890 --> 00:17:06,090 Josh: wrote and took a picture of. 298 00:17:06,470 --> 00:17:10,150 Josh: I didn't. It's GPT Image Gen 2.0, but it looks like it's handwritten. 299 00:17:10,250 --> 00:17:11,350 Josh: The handwriting was too nice. 300 00:17:11,450 --> 00:17:12,250 Ejaaz: Josh. That was the giveaway. 301 00:17:12,250 --> 00:17:15,730 Josh: Yeah, my handwriting is far sloppier than this. But the idea is that you can 302 00:17:15,730 --> 00:17:19,370 Josh: even write things on the back of a napkin and you could turn that into an application. 303 00:17:19,570 --> 00:17:23,230 Josh: So what we did here is we just asked for it to create a generic limitless dashboard 304 00:17:23,230 --> 00:17:27,590 Josh: application on the back of a piece of paper, fed it into the model, and this is what we got. 305 00:17:27,770 --> 00:17:29,930 Josh: So it looks like it did a pretty good job. 306 00:17:30,170 --> 00:17:33,970 Josh: I could tell this is Claude before you even tell me which model it is because 307 00:17:33,970 --> 00:17:35,630 Josh: it has the standard design principles. 308 00:17:36,150 --> 00:17:39,950 Josh: Claude design is so basic and 309 00:17:39,950 --> 00:17:42,670 Josh: it's so predictable where like okay i've seen this 310 00:17:42,670 --> 00:17:45,430 Josh: dashboard before it looks like it was a mission success there's a 311 00:17:45,430 --> 00:17:48,050 Josh: lot of text on this page a lot of stuff going on a lot 312 00:17:48,050 --> 00:17:51,370 Josh: of graphics i give a lot of credit for kind of inferring what 313 00:17:51,370 --> 00:17:54,090 Josh: we would want to be seeing from something like this where we have a 314 00:17:54,090 --> 00:17:59,050 Josh: proper trip budget i don't think we asked for a trip budget um but okay i think 315 00:17:59,050 --> 00:18:02,990 Josh: it looks like it made it did a lot of inferring right like it kind of made a 316 00:18:02,990 --> 00:18:05,250 Josh: lot of assumptions but in the end of the day it did take what we had on the 317 00:18:05,250 --> 00:18:10,910 Josh: napkin and it turned it into a pretty generic dashboard of sorts based on very 318 00:18:10,910 --> 00:18:12,130 Josh: limited information that we gave it. 319 00:18:12,830 --> 00:18:15,670 Ejaaz: I think the issue with this is we asked for something 320 00:18:15,670 --> 00:18:19,110 Ejaaz: completely different it created a dashboard um but 321 00:18:19,110 --> 00:18:22,170 Ejaaz: we asked it for it to be based around the limitless podcast and 322 00:18:22,170 --> 00:18:24,790 Ejaaz: it created a travel planning board so i don't know 323 00:18:24,790 --> 00:18:27,650 Ejaaz: whether that was a a prompt issue or whether we just fed 324 00:18:27,650 --> 00:18:30,510 Ejaaz: it the wrong image but but here we go here is where we're 325 00:18:30,510 --> 00:18:37,690 Ejaaz: at um now let's take a look at what openai did okay so here we have the same 326 00:18:37,690 --> 00:18:43,770 Ejaaz: prompt fed into gpt 5.5 and it's funny i can instantly tell this is GPT-515 327 00:18:43,770 --> 00:18:48,570 Ejaaz: because it's cleaner and it's not neon and it's not trying to go for some futuristic spin. 328 00:18:48,930 --> 00:18:53,590 Ejaaz: It looks very simplistic. This is actually a website or app that I would probably 329 00:18:53,590 --> 00:18:55,430 Ejaaz: be more inclined to engage with. 330 00:18:55,630 --> 00:18:58,110 Ejaaz: It's also more visually perceptive to me, right? 331 00:18:58,370 --> 00:19:03,490 Ejaaz: Like, what do I have at the front here? It's this five-day trip that I want to go on. 332 00:19:03,610 --> 00:19:06,210 Ejaaz: It's giving me the basic information that I need to know at the start. 333 00:19:06,210 --> 00:19:08,310 Ejaaz: It has a bunch of different tabs as well. 334 00:19:08,410 --> 00:19:12,110 Ejaaz: But again, it isn't what I specified on the napkin. So I think this might be 335 00:19:12,110 --> 00:19:15,390 Ejaaz: a skill to show on our side, Josh. But otherwise, like, look at these graphics. 336 00:19:15,610 --> 00:19:18,630 Ejaaz: They're like really good. One thing I've noticed is stylistically, 337 00:19:19,030 --> 00:19:26,010 Ejaaz: although both models create very different looking things, the animation style looks the same. 338 00:19:26,270 --> 00:19:29,250 Ejaaz: Have you noticed that even with the game previously that we just demoed, 339 00:19:29,370 --> 00:19:31,030 Ejaaz: the avatar looked the same. 340 00:19:31,170 --> 00:19:35,970 Ejaaz: It was given the same sort of title and the objects interacted in the same way 341 00:19:35,970 --> 00:19:41,670 Ejaaz: we're seeing this here so maybe it's just a change in quality i actually prefer gpt 5.5 on this one 342 00:19:41,670 --> 00:19:44,550 Josh: Yeah this is crazy i'm just going to suspect 343 00:19:44,550 --> 00:19:48,690 Josh: there was a prompt issue there where yes like we clearly we asked for something 344 00:19:48,690 --> 00:19:51,950 Josh: that we didn't actually want but here it is i think if you're just comparing 345 00:19:51,950 --> 00:19:56,850 Josh: them apples to apples uh chat gpt and codex is like no-brainer 10 times better 346 00:19:56,850 --> 00:20:01,010 Josh: i far prefer this if you look at the original napkin photo this is much more 347 00:20:01,010 --> 00:20:04,150 Josh: accurate to what the design looked like on that original piece of paper. 348 00:20:04,330 --> 00:20:07,410 Josh: And then if you also just compare the general design, this is far easier to understand. 349 00:20:07,670 --> 00:20:11,290 Josh: It's just a lot less dense. It's designed better. I wouldn't even say this is really... 350 00:20:12,190 --> 00:20:15,870 Josh: A fair comparison it seems like codex just like completely crushed this and 351 00:20:15,870 --> 00:20:19,770 Josh: it has all the functionality built in it looks good i am giving another win 352 00:20:19,770 --> 00:20:20,930 Josh: to codex here that's two for two. 353 00:20:20,930 --> 00:20:25,650 Ejaaz: Wow look i've got like a re-optimization uh toggle at the top and it actually 354 00:20:25,650 --> 00:20:29,050 Ejaaz: updated i wonder where it's pulling that data from it's already hooked 355 00:20:29,050 --> 00:20:31,130 Josh: Into data look at that yeah impressive stuff. 356 00:20:31,130 --> 00:20:36,210 Ejaaz: Very very cool now one major reason why both of these models have advanced so 357 00:20:36,210 --> 00:20:41,290 Ejaaz: rapidly over the last couple of months is something known as the ai model harness 358 00:20:41,290 --> 00:20:45,490 Ejaaz: now you have the AI model, which is something that you and I have interacted with quite a lot. 359 00:20:45,670 --> 00:20:48,210 Ejaaz: It's via ChatGPT or Claude itself. 360 00:20:48,450 --> 00:20:51,470 Ejaaz: But there's an added layer that you can put on top of this model, 361 00:20:51,590 --> 00:20:56,530 Ejaaz: which comes in the form of prescripted prompts that are engineered to make the 362 00:20:56,530 --> 00:20:58,570 Ejaaz: model act in a particular way. 363 00:20:58,690 --> 00:21:01,670 Ejaaz: But it's also the environment that the model works in. 364 00:21:01,730 --> 00:21:05,930 Ejaaz: It's also the policies that you set to make sure that the model acts and behaves 365 00:21:05,930 --> 00:21:08,310 Ejaaz: and sounds in a particular way. 366 00:21:08,410 --> 00:21:09,610 Ejaaz: That's why we talked about Claude's 367 00:21:09,610 --> 00:21:14,170 Ejaaz: personality earlier being better than ChatGPT. It all plays into the 368 00:21:15,970 --> 00:21:19,510 Ejaaz: We figured out was it's an entirely new product category on its own. 369 00:21:19,630 --> 00:21:25,050 Ejaaz: In fact, Cursor had some news over the last couple of days where they made their 370 00:21:25,050 --> 00:21:28,370 Ejaaz: harness, Cursor SDK, available via API. 371 00:21:28,490 --> 00:21:33,130 Ejaaz: And the reason why this is such a big deal is critics criticized Cursor for 372 00:21:33,130 --> 00:21:37,190 Ejaaz: being an AI wrapper, which meant that Cursor doesn't have a model of its own. 373 00:21:37,310 --> 00:21:44,970 Ejaaz: It would just create this harness, a set of prompts environments around, say, Claude or ChatGPT. 374 00:21:45,210 --> 00:21:49,430 Ejaaz: And so people would say, cursor isn't actually special. Turns out the wrapper 375 00:21:49,430 --> 00:21:52,670 Ejaaz: or the harness actually made these models way more intelligent. 376 00:21:52,810 --> 00:21:58,310 Ejaaz: In fact, if you added cursor's harness on top of GPT 5.5 and Claude Opus 4.7 377 00:21:58,310 --> 00:22:02,610 Ejaaz: right now, you end up with a smarter, more intelligent, more efficient model 378 00:22:02,610 --> 00:22:04,370 Ejaaz: than the actual base models themselves. 379 00:22:04,510 --> 00:22:07,790 Ejaaz: Now, remember, AI Labs spent hundreds of millions of dollars to train these 380 00:22:07,790 --> 00:22:10,350 Ejaaz: models and to create the best thing and put their best foot forward. 381 00:22:10,470 --> 00:22:13,030 Ejaaz: And still you have a startup which is worth, what is it now, 382 00:22:13,130 --> 00:22:16,750 Ejaaz: $10 billion right now, potentially being acquired by XAI for $60 billion, 383 00:22:17,270 --> 00:22:18,910 Ejaaz: creating a better model on top. 384 00:22:19,090 --> 00:22:23,170 Ejaaz: So the harness and the AI model are arguably one and the same at this point. 385 00:22:23,290 --> 00:22:26,290 Ejaaz: And it's just a valuable moat to point out that these models aren't just better 386 00:22:26,290 --> 00:22:30,670 Ejaaz: at coding because of the base model itself. It's because of this thing known as a harness. 387 00:22:31,110 --> 00:22:35,230 Josh: Yeah. And the harness is the difference maker when it comes to building this super app. 388 00:22:35,350 --> 00:22:38,190 Josh: It's like every single company is trying to build the super 389 00:22:38,190 --> 00:22:41,010 Josh: app the all-in-one application that kind of serves 390 00:22:41,010 --> 00:22:43,850 Josh: as your operating system anytime you need to engage with ai 391 00:22:43,850 --> 00:22:47,050 Josh: this is the place that you could do it and it's all encompassing it's 392 00:22:47,050 --> 00:22:50,090 Josh: all in one now one of the best applications we've seen for this in the early 393 00:22:50,090 --> 00:22:53,930 Josh: days has been something like open claw where it's this extension of what an 394 00:22:53,930 --> 00:22:58,850 Josh: operating system could look like starting with ai at the foundation and open 395 00:22:58,850 --> 00:23:02,170 Josh: claw did a really amazing job of that now in some news this week you can now 396 00:23:02,170 --> 00:23:05,990 Josh: use your chat gpt account to generate tokens with OpenClaw. 397 00:23:06,110 --> 00:23:09,810 Josh: So previously you had to use the API, whether you were using Anthropic or OpenAI 398 00:23:09,810 --> 00:23:13,150 Josh: or any of the other models, and it was pretty expensive. It costs a lot of money. 399 00:23:13,390 --> 00:23:16,890 Josh: Now, thanks to Sam Altman this week announcing, you can actually use your account 400 00:23:16,890 --> 00:23:20,270 Josh: connected with it. And I think this is the beginning of a multi-step plan. 401 00:23:20,800 --> 00:23:25,280 Josh: To really integrate OpenClaw directly into Codex in a way that Anthropik can't. 402 00:23:25,360 --> 00:23:28,200 Josh: Because if you'll remember, OpenAI owns OpenClaw. 403 00:23:28,380 --> 00:23:31,120 Josh: They bought Peter and granted OpenClaw will stay open source forever, 404 00:23:31,360 --> 00:23:34,460 Josh: but they have the ability to actually integrate directly into their products. 405 00:23:34,580 --> 00:23:35,920 Josh: And I suspect that's what we're going to see. 406 00:23:36,180 --> 00:23:39,960 Josh: In fact, we even got some confirmation from another post from one of the Codex 407 00:23:39,960 --> 00:23:44,460 Josh: developers who replied to a post that was saying, Codex only needs a native 408 00:23:44,460 --> 00:23:47,020 Josh: editor, an iOS app, a full browser, and OpenClaw. 409 00:23:47,180 --> 00:23:50,060 Josh: And the developer, Tebow, said all of this and more 410 00:23:50,060 --> 00:23:52,900 Josh: is coming to which sam altman retweeted it so we are 411 00:23:52,900 --> 00:23:55,580 Josh: indeed getting open claw inside of codex we're getting a 412 00:23:55,580 --> 00:23:58,500 Josh: mobile ios apps that you can access it remotely and soon 413 00:23:58,500 --> 00:24:01,280 Josh: there's going to be no reason to really use a different app 414 00:24:01,280 --> 00:24:04,680 Josh: because it's going to be all-encompassing now are there still downfalls yes 415 00:24:04,680 --> 00:24:08,820 Josh: computer use 20 faster on codex but yesterday i was playing around with it i 416 00:24:08,820 --> 00:24:13,560 Josh: told it to increase the volume of my music and it took 10 minutes to do it because 417 00:24:13,560 --> 00:24:17,200 Josh: it tried to increase the slider on spotify even though it was max without actually 418 00:24:17,200 --> 00:24:21,200 Josh: increasing my system audio so it's still a little dumb, but it is getting better. 419 00:24:21,360 --> 00:24:25,440 Josh: And I think this leads me to this post that I really love, the vanilla maxing 420 00:24:25,440 --> 00:24:27,000 Josh: post we have to talk about. 421 00:24:27,500 --> 00:24:32,140 Josh: Which starts by saying, you should 100% be vanilla maxing. Just use the tools 422 00:24:32,140 --> 00:24:34,000 Josh: as they're handed to you. That's it. 423 00:24:34,550 --> 00:24:37,110 Josh: Because a lot of people, and I've found this personally, and in fact, 424 00:24:37,230 --> 00:24:40,430 Josh: I've been caught by this personally, is that you try to get caught up in using 425 00:24:40,430 --> 00:24:42,270 Josh: all these different repos and these skills and these plugins, 426 00:24:42,470 --> 00:24:46,530 Josh: when the reality is, is if you just wait, the AI labs are shipping fast enough, 427 00:24:46,630 --> 00:24:48,610 Josh: they'll just integrate it into your own native application. 428 00:24:48,850 --> 00:24:50,810 Josh: So I'm vanilla maxing you, Jess. 429 00:24:50,970 --> 00:24:54,870 Ejaaz: I'm totally vanilla maxing as well, dude. Like, listen, OpenClaw, 430 00:24:55,130 --> 00:25:00,010 Ejaaz: when it was hyped up, was incredibly impressive and still is incredibly impressive. 431 00:25:00,170 --> 00:25:05,030 Ejaaz: It opened up an entirely new product market and segment. That's why OpenAI acquired them. 432 00:25:05,190 --> 00:25:08,430 Ejaaz: But something's majorly changed over the last couple of months, 433 00:25:08,550 --> 00:25:12,130 Ejaaz: which is OpenClaw has kind of fallen off. No one talks about it anymore. 434 00:25:12,450 --> 00:25:16,090 Ejaaz: People who are complaining about the errors and bugs that we're facing have 435 00:25:16,090 --> 00:25:18,710 Ejaaz: kind of gone silent because they've just grown bored and they don't want to 436 00:25:18,710 --> 00:25:20,030 Ejaaz: put their energy and effort into it. 437 00:25:20,150 --> 00:25:24,470 Ejaaz: And the reason why is because although these tools are very frontier level, 438 00:25:24,710 --> 00:25:27,570 Ejaaz: they can't actually be scaled to a practical use. 439 00:25:27,650 --> 00:25:31,410 Ejaaz: You don't feel safe integrating OpenClaw into your desktop where you have personal 440 00:25:31,410 --> 00:25:35,510 Ejaaz: files. I've seen horror stories where they access credit card data and expose 441 00:25:35,510 --> 00:25:39,530 Ejaaz: that or where they deleted old wedding photos and the wife was super angry, 442 00:25:39,830 --> 00:25:40,690 Ejaaz: a bunch of the stuff like that. 443 00:25:40,870 --> 00:25:45,170 Ejaaz: If you are able to get given or access to a tool that comes under a branded 444 00:25:45,170 --> 00:25:52,270 Ejaaz: reputation, such as ChatGPT, Codex, or Claude Cowork, where it kind of like 445 00:25:52,270 --> 00:25:54,970 Ejaaz: takes over your computer, but in a sandboxed environment. 446 00:25:54,970 --> 00:25:58,510 Ejaaz: I know that NVIDIA also released NemoClaw, which is like the enterprise-grade 447 00:25:58,510 --> 00:26:00,590 Ejaaz: secure version of OpenClaw. 448 00:26:00,890 --> 00:26:03,870 Ejaaz: You're vanilla maxing. That is the way to do it. And there's no need to rush 449 00:26:03,870 --> 00:26:09,230 Ejaaz: ahead and lose all your data as a consequence. So that's basically it for the episode. 450 00:26:09,410 --> 00:26:15,970 Ejaaz: We wanted to give you a comprehensive guide and insight into Codex GPT 5.5 versus Claude Opus 4.7. 451 00:26:16,090 --> 00:26:18,970 Ejaaz: There's a lot of numbers in there, but basically the best coding models from 452 00:26:18,970 --> 00:26:20,830 Ejaaz: both sides to see which is better. 453 00:26:21,050 --> 00:26:25,970 Ejaaz: And the truth is, there isn't a clear winner right now. I would say it's probably 454 00:26:25,970 --> 00:26:31,210 Ejaaz: Codex GPT 5.5, but the narrative switched so recently that maybe, 455 00:26:31,430 --> 00:26:34,870 Ejaaz: maybe Claude can still catch up. And the only reason why I say that, 456 00:26:34,950 --> 00:26:35,730 Ejaaz: Josh, is that's the only reason 457 00:26:36,020 --> 00:26:39,600 Ejaaz: there's a model that we haven't discussed or demonstrated yet because we can't. 458 00:26:39,680 --> 00:26:40,960 Ejaaz: It's called Claude Mythos. 459 00:26:41,300 --> 00:26:45,620 Ejaaz: It was kind of pseudo-released about a few weeks ago. 460 00:26:45,800 --> 00:26:49,900 Ejaaz: And on all benchmarks, it is technically better than 5.5. 461 00:26:50,080 --> 00:26:53,240 Ejaaz: But the reason why we can't demo it is we can't get access to it. 462 00:26:53,340 --> 00:26:56,900 Ejaaz: And the reason cited by Anthropic was because it's too dangerous. 463 00:26:57,060 --> 00:26:59,980 Ejaaz: It's a cybersecurity risk. In fact, it wasn't just Anthropic saying it. 464 00:27:00,120 --> 00:27:03,400 Ejaaz: It was Peter Heskett of the US Department of War also saying this, 465 00:27:03,540 --> 00:27:05,160 Ejaaz: right? So there's concerns around that. 466 00:27:05,460 --> 00:27:09,060 Ejaaz: OpenAI has created a mythos level type model here, but has made it available 467 00:27:09,060 --> 00:27:12,220 Ejaaz: to everyone. And so the argument could be made that it's just because Anthropic 468 00:27:12,220 --> 00:27:13,220 Ejaaz: doesn't have enough compute. 469 00:27:13,360 --> 00:27:16,200 Ejaaz: So there's a lot of rumors around this, but I'm excited to get my hands on the 470 00:27:16,200 --> 00:27:18,280 Ejaaz: best models from each of these and compare them directly. 471 00:27:18,600 --> 00:27:21,360 Josh: Yeah. And the compute's actually been degrading. So I think I want to wrap this 472 00:27:21,360 --> 00:27:23,800 Josh: up on like, what do you actually currently use? 473 00:27:23,880 --> 00:27:26,780 Josh: What is the limitless production stack? How are we using these AI models? 474 00:27:26,920 --> 00:27:29,760 Josh: And for me, at least, it's not even close. I'm codex-pilled. 475 00:27:29,920 --> 00:27:35,480 Josh: I'm fully switched over. I am codex superior domination. It's going to be the month of codex. 476 00:27:35,760 --> 00:27:39,080 Josh: Maybe Anthropic will have a comeback, but that's not happening until at least 477 00:27:39,080 --> 00:27:41,980 Josh: June, July, because this month is codex month. 478 00:27:42,100 --> 00:27:46,160 Josh: So I've been using codex for basically everything, all of the difficult tasks that I need. 479 00:27:46,300 --> 00:27:50,900 Josh: What I have found is that GPT 5.5 as an LLM, as a language model, 480 00:27:51,000 --> 00:27:55,200 Josh: as a chatbot is a little bit inferior to Opus 4.7, which I believe to be the 481 00:27:55,200 --> 00:27:57,320 Josh: better model if you're just chatting with an AI. 482 00:27:57,580 --> 00:28:00,720 Josh: I like its personality it's warmer it's more precise it normally 483 00:28:00,720 --> 00:28:03,640 Josh: gets the idea of what i want so if i am building a complex 484 00:28:03,640 --> 00:28:06,940 Josh: project opus 4.7 is the orchestrator and 485 00:28:06,940 --> 00:28:11,720 Josh: codex is the actual implementer the executor of this code of this plan i've 486 00:28:11,720 --> 00:28:16,240 Josh: also noticed that opus 4.7 is a bit inferior to 4.6 at a few things and i think 487 00:28:16,240 --> 00:28:20,760 Josh: this is another piece of alpha here i actually use opus 4.6 whenever i'm doing 488 00:28:20,760 --> 00:28:25,160 Josh: anything relating to writing or word ingestion so one of the projects i've been doing recently. 489 00:28:25,900 --> 00:28:29,400 Josh: Is andre carpathy he created this like wiki for 490 00:28:29,400 --> 00:28:32,120 Josh: your own person where it ingests files and it kind of writes 491 00:28:32,120 --> 00:28:35,880 Josh: these summaries for you and it creates a personal knowledge wiki i use opus 492 00:28:35,880 --> 00:28:40,920 Josh: 4.6 exclusively for that because opus 4.7 i think is far inferior at summarizing 493 00:28:40,920 --> 00:28:44,420 Josh: and kind of rewriting these topics that i use in my obsidian so that's kind 494 00:28:44,420 --> 00:28:48,200 Josh: of my stack i use opus for llms codex for everything else it just what are you 495 00:28:48,200 --> 00:28:49,880 Josh: currently optimizing for what. 496 00:28:49,880 --> 00:28:52,940 Ejaaz: Are you planning so it's two things when i have 497 00:28:52,940 --> 00:28:56,180 Ejaaz: a uh my stack is actually way more diverse when 498 00:28:56,180 --> 00:28:59,460 Ejaaz: it comes to just like the research side of things only because i'm 499 00:28:59,460 --> 00:29:02,300 Ejaaz: using the ai that's like available readily wherever i am 500 00:29:02,300 --> 00:29:05,320 Ejaaz: right so if i'm on x a lot and i see breaking news i'm 501 00:29:05,320 --> 00:29:09,100 Ejaaz: just tapping grok because honestly it's a recent model i think it's like what 502 00:29:09,100 --> 00:29:13,380 Ejaaz: was it 4.3 at this point uh is actually pretty good and they have multiple agents 503 00:29:13,380 --> 00:29:17,120 Ejaaz: that are kind of like running at this right but for the core bulk of the work 504 00:29:17,120 --> 00:29:22,400 Ejaaz: i've started shifting towards gpt 5.5 for the research because 5.5 research 505 00:29:23,100 --> 00:29:26,760 Ejaaz: things for so much longer. And it has a much more in-depth discussion. 506 00:29:26,880 --> 00:29:32,380 Ejaaz: In fact, I tested it out today because I was curious about the AI power stack 507 00:29:32,380 --> 00:29:36,520 Ejaaz: and what stocks I should be investing in to get exposure to the power grid lines 508 00:29:36,520 --> 00:29:38,680 Ejaaz: that are currently constraining AI data centers, right? 509 00:29:38,740 --> 00:29:41,720 Ejaaz: And I was like, all right, I gave a detailed prompt to both Claude Opus 4.7 510 00:29:41,720 --> 00:29:46,480 Ejaaz: and 5.5 and 5.5 completely cooked 4.7. 511 00:29:46,620 --> 00:29:51,200 Ejaaz: And it gave good reasoning why, whereas 4.7 did not. 512 00:29:51,420 --> 00:29:56,400 Ejaaz: I had to ask it more question. So all in all, I think 5.5 is my preference right 513 00:29:56,400 --> 00:29:58,880 Ejaaz: now. I still use 4.7 because of the personality. 514 00:29:59,120 --> 00:30:02,620 Ejaaz: It's like less of an AI type of voice versus GPT 5.5. 515 00:30:02,760 --> 00:30:06,000 Ejaaz: But again, I feel like OpenAI is on a generational run right now, 516 00:30:06,020 --> 00:30:09,900 Ejaaz: and they might just kind of fix this in the next couple of hours at this point. 517 00:30:10,420 --> 00:30:13,140 Josh: Yeah, it's coming. It's coming quick. And I think now is a good time to kind 518 00:30:13,140 --> 00:30:15,260 Josh: of get familiar with Codex to understand the way it works. 519 00:30:15,400 --> 00:30:19,740 Josh: And as they implement these features, you'll be able to adopt them within the hour, within the day. 520 00:30:19,940 --> 00:30:23,640 Josh: It's pretty amazing. And it's been fun to just experiment. It's been fun to try something new. 521 00:30:23,940 --> 00:30:27,940 Josh: And it's, again, competition is just better for everyone. So the end winner 522 00:30:27,940 --> 00:30:31,360 Josh: of this is the user, because for as low as $20 a month, you get access to all 523 00:30:31,360 --> 00:30:33,440 Josh: this frontier intelligence, all these capabilities. 524 00:30:33,760 --> 00:30:38,580 Josh: And it's just, it's really been unbelievable to watch. So that is the comparison, Codex versus Opus. 525 00:30:38,720 --> 00:30:42,620 Josh: If you have not tried both of them, I encourage you to give it a try. 526 00:30:42,900 --> 00:30:45,460 Josh: Test the prompts against one another. If you have any type of work that you 527 00:30:45,460 --> 00:30:49,280 Josh: need, if you're working on a computer at all, chances are you can use AI to 528 00:30:49,280 --> 00:30:50,680 Josh: help you do your job even better. 529 00:30:50,780 --> 00:30:54,160 Josh: Or you could just use it to help you do hobbies and side projects that you've 530 00:30:54,160 --> 00:30:55,520 Josh: always wanted to do. So give it a try. 531 00:30:56,100 --> 00:30:58,860 Josh: Let us know your preference, codex, cloud code. Which one is it going to be? 532 00:30:59,720 --> 00:31:02,400 Josh: I think that's probably it for the episode. Thank you guys so much for watching. 533 00:31:02,540 --> 00:31:04,720 Josh: If you enjoyed it, please don't forget to share with your friends. 534 00:31:05,100 --> 00:31:08,740 Josh: Let them know which model they picked. And also don't forget to rate it five 535 00:31:08,740 --> 00:31:12,860 Josh: stars on your favorite podcast listening platform. Any final thoughts, EJS, before we go? 536 00:31:13,220 --> 00:31:16,520 Ejaaz: No, that's it. Thank you guys so much for listening and we'll see you on the next one.