I Tested ChatGPT As My Cofounder for a Week - Here's Everything I Learned

0:00Today on the AI Daily Brief, I used ChatGPT Pulse as a startup co-founder for a week. And here's what happened. The AI Daily Brief is a daily podcast and video about the most important news and discussions in AI.

0:18All right, friends, quick announcements before we dive in. First of all, thank you to today's sponsors, Super Intelligent, Robots and Pencils, Notion and Blitzy. To get an ad-free version of the show, you can go to patreon.com slash AI Daily Brief, or you can now subscribe directly on Apple Podcasts. It's the same price, three bucks a month. Spotify will be coming soon when I figure out a couple of technical difficulties. For sponsor and speaking opportunities, go to aidailybrief.ai. All that information is there as well. And with that, let's get into today's episode. This is kind of a big, fun review show, perfect for a weekend big thing, where I look at three of the products that have come out this week, ChachiPT Pulse, Sora 2, and Imagine from Claude, and talk about how they did and which have the biggest implications for the future.

0:58Welcome back to the AI Daily Brief. Today, we are actually grading not one, but three different products that have come out this week. We're looking at ChatGPT Pulse, Sora 2, and the wildly overshadowed, but potentially most significant of all, Imagine from Claude. With both Imagine and Sora, it's going to be a little bit more impressionistic, but with ChatGPT Pulse, I had a very specific goal in mind when I wanted to look into it. So for those of you who don't remember the announcement, ChatGPT Pulse is basically a daily curated newsletter for you made based on your previous chats that comes up every morning with a set of potentially ephemeral articles that you can add to your general chats and engage with just like you would any other chat on ChatGPT.

1:38When it was announced, Sam Altman tweeted, Today we're launching my favorite feature of ChatGPT so far called Pulse. It is initially available to pro subscribers. Pulse works for you overnight and keeps thinking about your interests, your connected data, your recent chats, and more. Every morning you get a custom-generated set of stuff you might be interested in. It performs super well if you tell ChatGPT more about what's important to you. Think of treating ChatGPT like a super competent personal assistant. Sometimes you ask for things you need in the moment, but if you share general preferences, it will do a good job for you proactively.

2:08And this is really the key question, how valuable it is, how much people want a shift from reactive to proactive. For his part, Sam says, this all points to what I believe is the future of ChatGPT, a shift from being all reactive to being significantly proactive. New CEO of Applications, Fiji Simo, said something similar. In her Pulse announcement post, she wrote, AI should do more than just answer questions. It should anticipate your needs and help you reach your goals. She even called her blog post about it, a new paradigm of proactive, steerable AI. And she basically says that this is an agent that becomes more like a human assistant.

2:42Explaining more, she writes, Pulse combines information from your chats, your feedback, and connected apps to deliver a focus set of updates designed to make your life easier. That might mean following up on a project, suggesting resources to advance a long-term goal, or reminding you about an upcoming commitment. Personally, Pulse has already helped me discover new emerging treatments for my health condition, recommend new painting techniques for my art practice, surface great weekend events for my family, and more. All right, so on the one hand, they are explicitly in all of these announcements indicating that this is both for work and for life.

3:11You keep hearing this phrase, help you reach your goals. At the same time, it felt kind of inevitable to me that mostly this was going to be about, one, curating news around people's interests, and two, making contextual recommendations for things like restaurants, travel, events. And certainly when you looked at the OpenAI staffers who talked about their early experiences with Pulse, a lot of it was restaurant discovery or events or just interest aggregation. And so here was the issue for me going in. First of all, I am long on record as saying I just do not believe in and am not interested in personally any of the sort of restaurant, local events, real-world recommendation style features of these assistants.

3:53I do not believe that that set of functionality is sufficiently useful to justify agents in AI. I just don't think that discovering those things is enough of a problem that it justifies a whole new paradigm. Now, perhaps if that comes along with other stuff, fine. And I'm also very willing to recognize that my personal preferences here may not be the mass, but I cringe every time I see a new agent platform demo themselves by booking a flight or making a restaurant reservation. And so when people are finding family events as the main benefit of this thing, my eyes start to glaze over. The second thing is, it does feel like Pulse can do a great job for lots of people around information curation, around their interests and their professional pursuits.

4:30The problem with that use case for me is that in many ways, I am the human that for all of you. I spend literally hours every day reading not just what is interesting, but every single possible thing about AI so that I can filter and curate and synthesize what I think is most important. Now, that's not Pulse's fault. And I actually think that that feature can be super useful for a lot of people. I just knew that it wasn't going to be useful for me. In fact, one of the first things I had to do was make sure to instruct Pulse to never, ever try to share news about AI or related topics with me because it was just going to be a waste of space.

5:04All of this led to my starting position with Pulse, which was that while I thought it would be great for a ton of people, in fact, maybe for the average person, I didn't think it was going to be great for me, but I wanted to try it because in general, I found that I seem to get a lot out of ChatGPT as a business tool that other people don't. Just in terms of how I use regular ChatGPT, I use it, yes, for a ton of obvious stuff, for helping write ad copy, comparing YouTube thumbnails, comparing post titles. I use it for document review, translation, all the normal stuff that people use ChatGPT for.

5:39But then what's a little bit different is that I use it extremely deeply from a strategic perspective. And I've done this ever since 03 came out. The jump between 01 and 03 was where this use case was really unlocked, where I spend a meaningful amount of time pretty much every day talking through and sifting through my own ideas, strategies for the future, for both the podcast and surrounding media endeavors, as well as for super intelligent. And what I really wanted to try to see was whether I could tune Pulse to actually support that type of behavior to basically surface every day a set of interesting strategic thoughts, basically continuing our conversation in between the times that I was actually prompting it about something specific.

6:18I spent a bunch of time at the beginning really dialing it in to make sure I got less of what I didn't want and more of what I did want. And initially it did pretty well. In the first few days, I got some ideas for new product offerings. I got some review and reflection on positioning for super intelligent and how we can head into 2026 owning the idea of AI planning as a platform. And it also picked up on some discussions that I have been having around extension products for the Daily Brief in ways that were valuable. So when one of the articles on Pulse is something valuable, you can save it and it just becomes a part of your chat.

6:49And so you can see here this example where one of the things that I'm considering for the Daily Brief preview alert, because this isn't something I've announced or even actually decided if I'm for sure doing. But TLDR, I think there is an interesting gap between information and entertainment on the one hand, which is, of course, what the podcast does now, and full-on learning and upskilling on the other. I think there's a big space between them. And in that gap, there's a lot of opportunity to take informational or educational type content, tweak it a little bit to make it more useful and practical for upskilling without it being all the way into the realm of coursework and training.

7:21The idea that I've been discussing is the idea of an enterprise edition, which basically would take the show that we have as it is now and optimize it for cross-organizational usage. So what you can see here is ChatGPT Pulse trying to take it and position and package it a little tighter. The positioning, it says, AIDB Enterprise Edition, your org's daily AI learning layer. It operationalizes AI literacy and decision-making with lightweight routines embedded in Slack and Teams, email and recurring meetings, so busy teams absorb what matters and act on it. The core bundle from a product perspective, a daily operators cut, which is a one-pager with three to five takeaways, roll-by-roll actions and discussion prompts, a two-minute daily audio cut down, basically taking the show and compressing it for people who are really busy, a weekly team briefing that has what the implications of the top stories are, broader discussion points, etc.

8:08And then it even had a bunch of expansion pack ideas, as well as how to suggest ROI, pricing, distribution ideas, etc. So this is valuable. This is something that I'm actually actively thinking about. And like I said, in the first few days, I was able to get some of this style of thing out of here, which wasn't necessarily totally novel. But like I said, it did extend what we had been talking about. I also have been fairly successful in getting it to suggest totally novel products or offerings. While the hit rate on those is obviously going to be lower, that is still something that I think is interesting to basically have a constant ongoing overnight idea generator where I can quickly reject an idea or save it for later with the little bookmarking tool that puts it into the normal chats.

8:49However, I started to run into a set of very consistent issues which have really limited my ability to use it for these purposes. The first we'll call context creep. So like I said at the beginning, in addition to my primary or most valued use case, which is all of the strategic stuff, I do use ChatGPT for a ton of other things. And unfortunately, Pulse doesn't know which of the things that I'm using ChatGPT for are more important or less important to me. On any given day then, it's going to have a whole bunch of stuff that I absolutely under no circumstances want it to use as fodder for Pulse.

9:26Every day, for example, I'm using it to write descriptions and titles for podcasts. I'm using it to compare YouTube thumbnails. I'm sometimes asking it about topics and how to frame things for the show. And there are often very sort of basic mechanical things too. For example, the other day I was trying to figure out how Spotify subscriptions work and whether there was a way to add an ad-free version of an episode or whether I needed to publish each episode individually as its own thing and then mark it as only for the subscription. and the next morning in Pulse, the top four articles were all about Spotify subscriptions and optimization strategy, and it's just stuff that I had absolutely no interest in being there.

10:00But again, Pulse had no idea how to differentiate which things mattered most to me. Now, yes, theoretically, it is steerable, and I had some success in reminding it not to discuss things like thumbnails or YouTube or Spotify optimization, but it still has every day, a ton of that style of stuff creep in. And so until I'm really able to almost rank-prioritize which of my use cases for ChatGPT matter, if it's drawing from all of it, there's going to be some amount of this context creep. The next problem that it has, which is sometimes a more broad problem of ChatGPT and an imperfection of the memory system right now, is context confusion.

10:36The big context that ChatGPT gets confused at with me is, of course, where the lines between Superintelligent and AI Daily Brief are. Many times when I'm talking about a novel product, as clear as I try to be that it is something that I'm thinking about in the AI Daily Brief context or the super intelligent context, it ends up blurring them together. And this problem seems to be worse in the context of Pulse. It would frequently blend two different conversations together into one, or in another circumstance, extract details from one type of conversation, and plant it in another where it had no business.

11:07So for example, it situated the product that I was exploring in the right way, but pulled pricing discussions that we had had from something totally different from a different domain. Now like I said, this is me rubbing up against the limits of the memory architecture right now, and I assume that this will get better over time. But this sort of context confusion seems to be amplified, like I said, in the context of proactivity versus reactivity. The last common problem, which is really the big one over the last few days, was what I'll call the nothing left to say problem. There are only so many times it can rehash and repurpose and try to extend strategic conversations we've already had.

11:41It turns out that many of those things really require me to give new inputs and provide new direction in the way that I'm thinking about them for those conversations to be relevant. And so where I was consuming almost all of the content over the weekend at the beginning of the week, by the middle of the week, I was just giving it a glance and finding much, much less relevant. And even when it was bringing back up relevant strategic conversations, a lot of it kind of just got repetitive. Which brings me back to this assertion that AI should do more than just answer questions, that it should anticipate your needs and help you reach your goals.

12:13The verdict on that continues to be out for me. I am not unsatisfied at all with ChatGPT as a strategic partner. To the extent that I am, it's about things like that memory paradigm, not how proactive versus reactive it is. In fact, what I found all week was that for it to be useful in this context, it really required me to be continually inputting new questions, new thoughts, new explorations. The point is, for my strategic use case, I do not need ChatGPT to be proactive. I do not need it to be ambient. I worry, frankly, for OpenAI that this assumption that it should anticipate your needs is a little bit of old world Web 2.0 attention capture logic seeping in.

12:57Does AI really need to anticipate your needs? Or if it did anticipate your needs, would that just be a better way to get people to return to it? Now, obviously, the people who are building ChatGPT want people to use the product. And for the first time ever, tech companies are having to ask, is there enough that's enough? Might we not want the attention battle to be totally zero sum? But these are big ongoing questions and they will not be answered just in the context of this show. So as for a grade, I'm going to give it a B minus. I went in being very skeptical that it was going to be something that I would habituate.

13:31where I find myself a week later is that a lot of that skepticism has been validated, although not necessarily for the reasons that I would have thought. And frankly, for some reasons that are potentially fixable through some combination of product and model development, as well as my own personal tinkering. I find it interesting enough that I imagine that I will probably on most days open it up to give it at least a quick skim, flag the pieces that are relevant for me and come back to them later. And that might be enough. In other words, Pulse might not need 100 % consumption where I sit down for 20 minutes and read every word for it to be a useful feature.

14:03I do think more broadly, there is this theme and question of what the big game is for all of these products. I really, really, if you are at OpenAI and you are listening, I would so strongly encourage your communications team and your leadership specifically, articulate your vision for how these attention consumptive experiences all add up to part of the master plan. Between Pulse, which feels so much to people like a vehicle for ads, the Sora app, which we're about to talk about, there is a new narrative that the new guy is turning into the old guy, and all it cares about is keeping our rapt attention for the sake of ad dollars.

14:40It strikes me that that is probably a reductive way to look at things. Sam Altman seems to have indicated as much, but if that is not the case, communicate it and let people know how it all ladders up to something bigger. So that's ChatGPT Pulse, but there were, of course, a bunch of other interesting products released this week. I'm not going to have number grades for them, but the two that I want to talk about are Sora 2 and Imagine from Claude. Today's episode is brought to you by Superintelligent. Now, one thing that we are having a lot of conversations with folks about is the fact that for some of you, your fiscal year is coming to an end, and that means two things.

15:14One, it means planning and thinking about what you're going to do in the next year. And two, it means using up those last of budgets so you don't lose them. If you are an enterprise that happens to find yourself in that situation, Super Intelligent would love to help on both fronts. We are moving increasingly towards an annual AI planning model where we map out how you can create an action map of your organization's agent opportunities that represents an executable backlog of AI and agent use cases that you can deliver on over the course of the next year. Additionally, for those end-of-year budgets, we have worked out deals with a number of partners where we can pre-lock in general implementation packages even before you've figured out exactly what use cases are going to require them.

15:53If you'd like to learn more about superintelligence agent readiness audits and this new end-of-fiscal year plan, visit us at bsuper.ai, click Get Started, and make sure to use the word fiscal somewhere in the description. AI changes fast. You need a partner built for the long game. Robots and pencils work side-by-side with organizations to turn AI ambition into real human impact. As an AWS-certified partner, they modernize infrastructure, design cloud-native systems, and apply AI to create business value. And their partnerships don't end at launch. As AI changes, Robots and Pencils stays by your side, so you keep pace.

16:27The difference is close partnership that builds value and compounds over time. Plus, with delivery centers across the U.S., Canada, Europe, and Latin America, clients get local expertise and global scale. For AI that delivers progress, not promises, visit robotsandpencils.com slash AI Daily Brief. Chatbots are great, but they can only take you so far. I've recently been testing Notion's new AI agents, and they are a very different type of experience. These are agents that actually complete entire workflows for you in your style, and best of all, they work in a channel that you already know and love because they are purpose-built Notion super users.

17:04Notion's new AI agents completely expands the range of what Notion can do. It can now build documents from your entire company's knowledge base, organize scattered information into organized reports, basically do tasks that used to take days, and get them complete in minutes. These agents don't just help with work, they finish it. Getting started with building on Notion is easier than ever. Notion agents are now your very own super user to help you onboard in minutes. Your AI teammates are ready to work. Try Notion AI for free at the link in our show notes. This episode is brought to you by Blitzy, the enterprise autonomous software development platform with infinite code context.

17:38Blitzy uses thousands of specialized AI agents that think for hours to understand enterprise-scale code bases with millions of lines of code. Enterprise engineering leaders start every development sprint with the Blitzy platform, bringing in their development requirements. The Blitzy platform provides a plan, then generates and precompiles code for each task. Blitzy delivers 80 % plus of the development work autonomously while providing a guide for the final 20 % of human development work required to complete the sprint. Public companies are achieving a 5x engineering velocity increase when incorporating Blitzy as their pre-IDE development tool, pairing it with their coding co-pilot of choice to bring an AI-native SDLC into their org.

18:14Blitzy is providing a limited-time, 30-day free proof-of-concept for qualifying enterprises. The team will provide a 5x velocity increase on a real development project in your org. Visit blitzy.com and press book demo to learn how Blitzy transforms your STLC from AI-assisted to AI-native. That's blitzy.com. Talking first about Sora, by which I really mean Sora 2, the model, and Sora, the application, I'm not going to rehash the endless debates around short-form video content and whether OpenAI should be doing this at all. I'm just going to look at it on its own terms. So first of all, let's talk about the things that people are doing with it.

18:52It is still very much in the exploration phase. The endless videos of Sam Altman from the first 24 hours have thankfully dissipated, but there are still a few common trends that you see over and over again. We're getting, for example, a lot of Pokemon and Pikachu. Bob Ross and dead celebrities are kind of everywhere. There are a bunch of historical remakes. I'm seeing MLK, JFK speeches remade. There are a lot of animals driving cars and getting pulled over. So what have I been playing around with? Well, there are three, I think, big themes. Nostalgia and historic generations. Seasonal content.

19:28It is October after all. My favorite combination of those is this one. Look, I got open AI. I got Google. And I got Anthropic. Let's team up. Yeah! This Halloween, every Happy Meal comes with one of three AI pals. By the way, part of why I think people are attracted to the nostalgia stuff is that in addition to that just being kind of where we are right now as a society, it also makes it easier for the model to perform well. If you go for something nostalgic, there's a higher chance, in other words, that the model output is going to match roughly the vibe or the emotion that you were trying to match.

20:00By and large, so far, Sora has been pretty much fun and playful rather than people thinking about the business use cases of it. But for Kix, I did start to experiment with, could you do micro-explainers using some of these same ideas? So here's, for example, one about context engineering. Hey, future thinker, ever wonder what makes AI answers smart and not just lucky? It's all about context engineering. Context is the who, what, where, and when that surrounds your data. In the search for thinking machines, researchers are turning to a method inspired by life itself, reinforcement learning. A system tries, fails, and tries again.

20:35But what about the cameo feature that was such a big part of the pitch, where you could put yourself in videos or allow other people to put you in videos. The jury's still out here. Certainly people have been playing around with it a bit, but at least for me right now, it feels like either novelty or I just don't have the social density for it to be a real mode of interaction. Honestly, I kind of get the feeling that we need to get this app in the hands of the youth as fast as humanly possible to understand if it's actually going to be a thing. I can totally see the ability to put you and your friends in videos as being so different and so fun that it actually does end up justifying its own digital space.

21:10But to some extent, that'll be the big question when it comes to Sora app longevity. Still, for now, the thing is number three in the app store. And so at least for now, interest remains high. And by the way, for those of you who are interested in some of the business use cases of Sora, count on Greg Eisenberg to always have a fast list of where money can be made. Some of his thoughts, he thinks that someone will buy a synthetic Sora 2 account for 10 million, and it'll be a good deal, but look crazy. He thinks that whoever builds the no AI Instagram will make a bajillion dollars because people are going to get sick of Sora spam.

21:40He thinks that people are going to start selling their own likeness, that someone's going to build a daily Sora channel and then flip it into a startup, and so on and so forth. Now, the last product I wanted to talk about is the one that is the least mature and potentially the long-term most disruptive. That product is Imagine with Claude, something that Anthropic is calling simply a temporary research preview. Imagine is basically an experience where Claude generates software on the fly. They write, no functionality is predetermined, no code is pre-written. What you see is Cloud creating in real time, responding and adapting to your requests as you interact.

22:11And Imagine is restricted to Anthropic's highest tier, so there's a very small handful of people who have it, but the results so far are pretty impressive. In a lot of ways, what it's showing is Sonnet 4.5's ability to create user interfaces. Let's do a couple of live examples. Here's one of their suggestions, make a time machine control panel. Let's click it, and we'll actually watch this in real time. I'll have my editors not fast forward this. You can see it working. And it's generating it in real time. So as opposed to a vibe coding tool like Lovable, where you give it your prompt and then it figures it out and writes it, this is all happening in real time and is actually interactable at the moment that you're first seeing it.

22:53Now, obviously, this is just an interface. It doesn't have a backend, at least not fully. So there are limits to what you can do. And this is where the research preview aspect of this comes in, but it's still pretty cool. One of the experiments that I like doing when it comes to these types of experiences is design a version of 1990s black and green Oregon Trail, but with lovecraft themes called Eldritch Trail. I like this because it's a distinct experience that it can pull back on with a distinct reframing that gives you an easy ability to see how well it does. For any of you that grew up in the late 80s and early 90s, you will recognize that we are in the right visual zone.