10 Ways to Think Bigger with Opportunity AI

When GPT-6 Astra was released, people quickly noticed that it was a little bit different than other previous model releases. In some ways, it was unbelievably more advanced than anything we'd seen before. And yet in others, not only were the improvements not necessarily super noticeable in certain use cases, but sometimes it actually felt like a regression. The interesting point in the history of the development of AI that we've come to is that broad model capability has reached a level where the value of new models is very frequently not going to be in just doing the same things that you've been doing with AI better, but actually about totally unlocking new capabilities that you've never even considered. That said, unlocking new capabilities that you've never considered by definition means you haven't considered them. And so how do you figure out even what is valuable to use that AI for? That is the goal of today's episode, to give you a set of thought starters about what I call use cases for opportunity AI. The AI Daily Brief is a daily podcast and video about the most important news and discussions in AI. All right, friends, quick announcements before we dive in. First of all, thank you to today's sponsors, Blitzy, Section, Robots and Pencils, and HyperAgent. To get an ad-free version of the show, go to patreon.com/AIDailyBrief, or you can subscribe on Apple Podcasts. And to learn more about sponsoring the show, send us a note at sponsors@aidailybrief.ai. Today we are talking about ways to use AI way better. And the specific context for this exploration is around the latest generation of models, Fable 5.1, and especially GPT-6 Astra. Astra is in some ways a really strange model. Like I said in my show, I called it so significant but also so confounding. And over the last week since it's been available, you can kind of see examples of this if you're paying attention to the advanced AI users. For example, a couple of days ago, Francesco on X wrote, moved back to GPT-5.6 SOL and Fable 5.1. GPT-6 Astra might be the smartest and dumbest model I've ever worked with. I can't deal with its mood swings, especially the absurd shortcuts it takes to arrive at something technically working. Theo reposted that and said, remember when everyone said I was overreacting to Astra's spikes of stupidity? Meanwhile, OpenCode's Dax wrote, A portion of our team has gone back to SOL. Astra is good and can do some novel things, but it has some downsides. And so far our effective spend looks doubled, so tough to justify. Now all of these folks are dealing primarily with the coding use case, but that's not necessarily where a lot of the initial excitement about Astra has been. Instead, the excitement about Astra has been around some totally new capabilities that get unlocked around things like video editing and 3D design and modeling, things in other words that are not currently part of our day-to-day work. Now, whether that's a good strategy for OpenAI from a business perspective is a whole separate question, but the larger concept that this gets into is one of my most frequently discussed ideas, the difference between efficiency AI and opportunity AI. Efficiency AI is of course AI that helps you do your existing work better. Better can mean faster, more efficiently, more cheaply. Opportunity AI, on the other hand, is AI that unlocks entirely new opportunities. Now, efficiency versus opportunity AI, in most times that I've discussed it, has been more about a mindset for how to use and think about AI than some actual specific difference in the models. With Astra, it's one of the first times that I've seen a model that actually dances inside this difference. Now, it's important to note that there is absolutely nothing wrong with efficiency AI. It is going to be the foundation for most of our use of AI. It's where a lot of the initial value for AI is going to come. The reason that I've always discussed opportunity AI is that I think especially on a business strategic level, if companies only think about AI as an efficiency technology, they're going to miss a lot of the opportunities that the companies that ultimately win will not. In other words, all of those efficiency use cases will become table stakes and will reset the expectations of how work gets done. But the companies that lean into finding and discovering new opportunities, even if those opportunities are currently orthogonal to what they do, are likely to be the ones that race out ahead and really transform themselves for this new era. So where I left the Astra conversation a few days ago was that it was all about opportunity AI, and I guarantee that some of you were sitting there thinking, Well, that's well and good, but what the heck does it even mean to use this model to unlock entirely new opportunities? Maybe some number of us have some things that we'd like to do that we've never been able to before, but I have found that certainly for myself, a lot of what I would categorize as my opportunity AI use cases, I kind of had to stumble and bumble into. In other words, it wasn't like I was sitting around just waiting for the technology unlock to build a new type of website that could automatically turn these episodes into shareable chunks and then push them into a pipeline that could provide social and video, that came after a lot of stumbling around and experimenting and looking at problems that I had and trying to solve them in new ways. And I think that if you just ask people in general to go use AI to find new opportunities, you're gonna have a pretty massive blank page problem. In short, people do not generally walk around with some complete inventory of all the things they might do or that they might make. We have a smaller, more familiar inventory shaped by our job, our tools, our experience, and what we see the people around us doing. So with that in mind, what I wanted to do for this long read slash big think episode was try to actually explore and expand that whole new world of possibilities. I'll be adding the web experience that I'm now talking over to the aidailybrief.ai website so that you can check it out as well. And basically this is 2 parts. The first, is an exhibition of 12 ideas of things that you could build or do or use a model like GPT-6 Astra for right now that you might not have thought of. They are, in other words, thought starters. Each of them has both a high-level concept as well as a specific application of that concept. And for each, there is a little personalizer where you can tell the AI a little bit about yourself and find some similar possibilities that maybe better fit your life and work. Now, one shortcut, if you don't want to go through all of these with me, is that a lot of the quote unquote opportunities of opportunity AI are not totally novel new things you can do, but things that maybe you specifically haven't been able to do, but other people have been able to do. In other words, a good shorthand for looking at different ways to use AI that may stretch you even farther is to look around at what people with other jobs are doing that you think is really cool. You'll see as we get into it what I mean. So the first thought starter is to make something that people can play, specifically marketing that people can play. Historically, the content for marketing has been some combination of visual and print and more recently video. But AI creates a whole new opportunity around interaction and interactivity. Some of the most exciting first experiments that people did with Astro were to create games or to rebuild games that already existed. And so if that is a capability of something like Astra, why not think about where building games could be useful inside of work? You could invite a person to explore a world, take on a role, make something, attempt a challenge. Their relationship to the brand or the product develops through what they do in that situation. This is a vast creative territory, in many ways larger than deciding which message to put on a page. Consider the difference between being told a place rewards curiosity and being given some small mystery that makes you curious about the place. In the first case, curiosity is the subject of the message, but in the second, curiosity is something the experience asks you to exercise. As opposed to other approaches to marketing where the recipient is purely a receiver, creating a game implicates the audience's agency. And there is a huge variety of what it could mean to integrate a game with the branded messages. In one example, the rules of the game could carry the argument for the product or service. Suppose you advise growing businesses and believe that coordination becomes a bigger problem as teams expand. A short game could ask someone to deliver a project while adding people and managing handoffs. If adding capacity also creates more coordination work, the player encounters the relationship your advice is built around. In another type of game, play could help someone discover why a problem matters to them. Imagine a consultant whose services sound abstract in a sales deck. Improving cross-team decision-making. With a game, you could give a prospect a 5-minute fictional launch. Sales promises a date, product discovers a dependency, and support needs information nobody has assembled. The prospect makes choices and encounters the resulting confusion. The experience can give them language for a problem they recognize from work. In another instance, play can let people express who they are. Think about a make this awkward apartment work challenge. The player has a room, a few competing needs, and limited space. They move pieces, choose what deserves to be highlighted, and arrive at a particular solution. The experience places the products inside a problem the customer can understand and lets the customer exercise taste. Now, this is not something that hasn't been done before. Brands have experimented with games as marketing. It's just that in the past, these things were extremely constrained by the sort of development resources that it took to make them, as opposed to something that you literally could this weekend experiment with As a solopreneur. And the last thing I'll note on this concept is that it also highlights the fact that especially when you're dealing with opportunity AI, not everything is going to hit. Game design is of course about more than just being able to tell a coding agent what to build. There's a reason that a lot more games are released than ever become popular, and pretty much all of those, even the ones that fail, are created by professional game designers. So without at all minimizing the challenge, of building a compelling experience, the fun thing is that these new models and opportunity AI mindsets allow you to actually try. Second thought starter is once again taking something that some people do and bringing it into your own work world. One of the things that these new models are getting very good at is building video production pipelines. If you have ever seen any clips from the AI Daily Brief, those are not created by a dedicated video clipping product like Opus. That is a custom-built Cloud Run pipeline that ingests script, the raw video material, and then uses a set of tools to produce the videos. And to be clear, I think that I am barely scratching the surface on what that pipeline can do. I have seen people doing some amazing things with video editing with Astra, and I wouldn't be surprised if these more advanced models have a similar sort of democratization effect on video production as the coding agents have had with building software. The interesting question becomes, where could video help you in your work? Is it about explaining something either internally to employees or externally to customers and potential customers? Would video be more valuable in the marketing realm? Could you combine an educational impulse with the new video capabilities of advanced models to produce training materials that also serve as marketing? And before you get hung up on the video capture part itself, we all have laptops and phones at this point. That at the very least could capture us talking into them. So here's the homework I'll give you if you wanna try this out. Start by doing something simple like an educational marketing video for whatever it is that you build or sell. Work with AI to write a 60-second script and record it in the simplest way possible just with your iPhone. Then give that video either to Codex or Claude Code and ask it to design a visual motif for videos that you do. Ask it to design transition elements or layered graphics and tell it you want a full production pipeline so that all you have to do is drop this source video in and it can take it from there. Whatever it comes back with, give it at least one round of feedback and see if you can push it farther. And then sit back and ask, is this sort of video creation something that you could now actively consider in a way that you never have before? In other words, if your contribution to releasing video comes down to initiating script production, and recording yourself reading that script, does that sufficiently lower the barrier to entry where video could become a tool in your work? And from there, you can take video in a lot of different directions and go a lot farther. I would also suggest that if you are interested in experimenting with this sort of capability, but can't think of any place that it would be super useful in your work, go partner with an AI to write a little 90-second movie and try to build a production pipeline and strategy from there. Every AI coding tool on the market does the same thing first: it starts writing code. Blitzy does the opposite. Before writing a single line, Blitzy spends days reverse engineering your entire codebase. Thousands of agents ingest millions of lines, mapping every dependency, every undocumented constraint, every architectural decision made over the last decade. The result is a dynamic knowledge graph that understands your software the way a principal engineer would after 30 years in the building. Other tools guess at context with grep searches and markdown files. Blitzy never guesses. It builds true understanding first, then delivers over 80% of entire software epics autonomously. Validated, end-to-end tested, production-grade pull requests. That's why Fortune 500 engineering teams trust Blitzy with the codebases that matter most. See for yourself at blitzy.com. That's B-L-I-T-Z-Y dot com. Here's a harsh truth: your company is probably spending thousands or millions of dollars on AI tools that are being massively underutilized. Half of companies have AI tools, but only 12% use them for business value. Most employees are still using AI to summarize meeting notes. If you're the one responsible for AI adoption at your company, you need Section. Section is a platform that helps you manage AI transformation across your entire organization. It coaches employees on real use cases, tracks who's using AI for business impact, and shows you exactly where AI is and isn't creating value. The result? You go from rolling out tools to driving measurable AI value. Your employees move from meeting summaries to solving actual business problems. And you can prove the ROI. Stop guessing if your AI investment is working. Check out Section at sectionai.com. That's S-E-C-T-I-O-N-A-I dot com. The best teams don't have a single star carrying everyone else. They know their own strengths and each other's weaknesses and play to both. That's the team Robots and Pencils has built on purpose. Nobody there is grinding through busywork to pad a headcount number. People come for the hard problems, and they stay because everyone around them is leveling up at the same time. In a market full of companies that are just trying to hire fast, that's worth a look. Check out robotsandpencils.com/careers. This episode of the AI Daily Brief is brought to you by HyperAgent, where you run fleets of agents your team can manage together. Forget local agents and chat workflows waiting on your laptop to be prompted. HyperAgent deploys always-on agents in the cloud, doing real work across the tools your team already uses. Marketing agents turn competitor moves into landing pages. Sales agents enrich leads, draft emails, and updates the CRM. Ops agent chases the paperwork and tracks the budget. Every agent has access to shared context and follows your rules about scope and approvals. It's time you had agents that feel like teammates. Hire yours at HyperAgent. Get $100 in credits at hyperagent.com/AIDailyBrief. Third thought starter idea: product demos that people can explore. A central new possibility is giving a customer control over how they come to understand a product. A product contains many potential explanations, and different buyers arrive with different questions. An interactive representation can let the question determine the path. Right now, for example, a sales presentation has to choose an order. A buyer concerned about maintenance may sit through features they already understand before reaching the part that matters to them. A buyer concerned about configuration may need to see a relationship that the standard sequence never shows. An exploratory demo can organize itself around actions: open this, isolate that component, check the arrangement, inspect the result. The visitor's curiosity becomes a way to navigate the explanation. The question to ask is, what do your customers need to be able to inspect for themselves before your product or service really makes sense? And is that something that an interactive demo could solve? Thought starter 4 comes from a similar place and is proposals that clients can shape. Right now, proposals point in a single direction. The proposer hands the proposed to a plan and hopes they like it. However, there's no reason anymore that a proposal can't become a shared instrument for making a decision. It can contain enough of the relationship between scope, resources, and time that clients can explore alternatives instead of receiving only one selected plan. Think about it this way. A proposal already contains a model, even when that model is invisible. The author has made assumptions about how much work is involved, what can happen in parallel, which resources are available, and how changing the brief affects delivery. When a client asks, could we include another department, the author runs part of that model again. The answer may involve more cohorts, another facilitator, or a later completion date. And so instead, can't you make selected parts of that reasoning usable by the client? The opportunity is to make selected parts of the reasoning in your model usable by the client. An interactive proposal could show what would follow from changing a part of the plan. So for example, we want it sooner becomes we can finish sooner if the team can attend more often, or we need more capacity to run these activities in parallel. The client could potentially discover that their initial preference conflicts with something they care about more. They can explore that conflict without treating every alternative as a new request for you to interpret. Now, what's fascinating about this one is that if it's done well, this is actually efficiency AI and opportunity AI all in one. Not only does this represent a new way of interacting with clients that wasn't really possible before, but in so doing, it radically reduces the latency of the back and forth that is a necessary part of every new client negotiation and has the feel to me of the type of thing that in the future is going to be completely de rigueur. And we're going to find it hard to remember a period where we didn't do things in this sort of interactive way. For our 5th thought starter, we're turning our focus to the internal. To me, a primary category of AI value is allowing us to think about strategic decisions from different angles. AI's sheer ability to absorb information and output lots of possibilities adds a real dimension to strategic decision-making. And so the 5th idea is to build simulators for business decisions. For what it's worth, this also gets to some of the concepts in our multiplayer AI sprint, which is all about the agents that will exist at the center of teams rather than just owned by individuals. So the idea of the business decision simulator is that many disagreements actually contain several disagreements hidden inside. We need another hire could mean that the current process is too slow, the work is unevenly distributed, or that someone expects some future surge in demand. What that means is that people end up arguing about the conclusion while imagining different starting conditions. One person thinks about the average week, another remembers the worst day, a third assumes the process will improve next month. Constructing a simulator forces you to specify those conditions. You have to decide what happens at each stage, what limits each stage, where unfinished work goes, things like that. And a simulator can make the consequence of an assumption something that can be inspected. You can ask whether the same decision still makes sense if demand rises more slowly, or if training a new employee takes time, or if one stage handles more complicated cases. And in this, you'll notice another pattern that started to weave itself throughout a lot of these ideas that many of them come back to in some way— building what-if machines. That was sort of the idea of the interactive proposal, right? What if we did this? What would the implications be? What tradeoffs would it implicate? This is a what-if machine, but in the context of some specific decision. Our 6th thought starter, I'm actually going to skim over. It will be on the website for you to check out, but the specific example is going to be perhaps a little bit more narrow. And so the thing that I wanted to touch on is just that one of the clear capabilities that particularly Astra exhibits is the ability to operate in 3D. A lot of the first examples of people really being excited about Astra were them using it with 3D software like Blender to do things like 3D walkthroughs of Zillow houses that they might want to buy. Or learning experiences where 3 dimensions can make a big difference. I do think that one of the best ways to think about opportunity AI when it comes to Astra is to ask where 3 dimensions could be transformative in some way. And one area that I would suggest looking to are learning experiences where being able to interact with and rotate some 3D digital object can make a big difference in how you learn. Some other ways to think about 3D are of course going to be marketing use cases, different types of videos that involve 3D generation. Honestly, I think at some point this might be an entire show all on its own. For now, the big takeaway is that 3D is a big part of what makes Astro unique and where it might be worth spending some of that opportunity time. A 7th thought starter is another instance of the video editing pipeline capabilities, and the proposal is to turn customer stories into films. Organizations often already possess the raw material for this. Think about an interview with a customer, a project review, some recordings, screenshots, photographs. Those are things that with a model like Fable-5-1 or Astra could be turned into a mini documentary. And this is more than just using video for video's sake. Narratives help bring evidence to life. A screenshot of a finished workflow may show what exists, but it doesn't on its own explain why anyone needed it. An interview may explain why the situation mattered, but the viewer still needs to see what changed. You probably get where I'm going with this. You take those things and put them together in combination, And it answers both questions. Imagine a customer describing a failed handoff. The speaker explains what information was missing and what prevented them from doing. A recording of the new process then shows where that information now goes. The outcome has become specific enough to understand. The strength is in the relationship between those moments. The film recognizes the audience to recognize the original problem, inspect the invention, and evaluate whether the outcome addresses the problem described. Film also creates an expanded canvas to show judgment. What alternatives were considered? Why was one decision difficult? What did the team learn after the first attempt? Very similar to the previous video example, my suggestion for this one would be to try to take the raw assets of a customer journey, dump them into either Claude Code with Fable 5.1 or Codex with Astra, and at least for the sake of the first time, ask it to architect and put everything together to see what it does in one shot. Now, I don't think one-shotting these things is the actual way that you're going to want to do this in the long run. But I think it's going to be easier to understand the value of this once you're actually seeing and interacting with it. And so don't take too much time in advance to get it perfect before you have a sense of the capability more generally. Thought Starter 8 is one that I think is really interesting and exciting because it's basically a way to extend education and learning and professional development in a powerful new way. The core idea is to create an experience or a simulation of the experience before the real situation occurs. It could be a place, for example, for people to practice how they handle difficult customer situations, Or difficult communication practices with management. One of the things that's hardest about learning experiences is that it's hard to create a lot of space for learners to have a chance to express and get feedback on judgment. Practice environments can be really good for exactly that. And by creating a simulation-style environment, you have the possibility of getting immediate feedback on that judgment in a way that allows you to iterate and navigate it midstream. This is another one where I actually think this is going to be a whole category of professional development experiences that share this broad root in practice environments and simulation environments for internal learning. And I'm super excited to see what people create in this front. Thought starter number 9 is good either A, for product designers who actually design physical products in the real world, or B, for people who want a way to experience the power of something like Astro with 3D modeling, but who don't necessarily personally have a use case for exactly that sort of capability. The thought starter is about physical products that you can prototype, and the idea is to treat physical conditions as something you can design. Now, this could be something like an actual product, but it could also be a little bit more abstract. For example, a teacher might create an object that has removable pieces to make some difficult relationship they're trying to explain more tangible. And what I'll say here, 'cause this probably feels pretty abstract to a lot of you, is that this is exactly why I built in this Make It Mine feature. I am absolutely not promising it will be perfect, But if you give it a little bit more context about you, for example, I'm building a new type of podcast studio and interested to think about where this sort of prototypable physical products might fit into how I work or what I'm going to need to do. You give it that info and then you ask it to find possibilities. So in this case, the 3 ideas it came back to are turning your episode rundown into sliding blocks guests can touch. Not really a fit for me, but certainly creative. Number 2, let the room's own measurements shape your acoustic panels. That one has more promise as something that might actually be valuable. And building your studio as a hand-sized kit clients rearrange. My fidgety kids might like that, although I'm not sure that that's particularly relevant for me, but you get the idea. The Make This Mine panel is hopefully going to give you some nuggets of how you can, well, make these ideas yours. Now we're running out of time, so I'm actually going to skip Thought Starter 10, Building Browser Features, and Thought Starter 11, A Test Crew for Your Website. But as I mentioned, all of this will be available on aidailybrief.ai for you to check out. And we'll close with the last one because I think it's something that might be a lot more relevant for a lot of you knowledge workers, especially given how many of you I know are in some form of the business of helping clients directly with specific types of intellectual and knowledge work challenges. So the idea behind this last opportunity AI thought starter is your expertise as a product, basically giving a specific part of your judgment a form someone can work through. Now, of course, expert help tends to manifest itself as a conversation, but several kinds of work are happening inside that. You are gathering context, recognizing patterns, noticing exceptions, ruling out attractive but inappropriate options, and deciding what the person is ready to do next. The visible advice is the end of that process, but there's an interesting question about whether any part of that could actually turn into a product that people could manipulate and interact with themselves. To get just a little bit meta, This entire interactive web app that I'm publishing alongside this episode is sort of an example of this, right? Instead of me sitting with each of you, giving you my best ideas for what opportunity AI ideas you might want to go pursue, I've embedded a lot of that in this interactive experience that while sure isn't the same as sitting and having my undivided attention, gives a lot of people a pretty good chunk of the value that might have come from that sort of conversation. By turning our expertise into products, we can scale ourselves. And by the way, this doesn't just have to apply to people who have clients. You also might be able to turn your expertise into a product that's available to other members of your team or your broader company and organization as well. And fascinatingly, once again, we close on a thought starter that even though the building of this is firmly in the opportunity AI camp, because most of us have never sat around asking whether we could build a digital advisor version of ourselves, in a lot of situations, this will amount to efficiency AI as well, especially inside those company contexts where a lot of our time is spent on repeating the same things to different people. So to close out again, efficiency AI and opportunity AI are not somehow locked in mortal conflict. The goal of opportunity AI is simply to stretch ourselves and ask what we might do now that we never would've before. This episode hopefully has given you some ideas that won't have required you in advance to know exactly what opportunities could look like for you, and I'm excited to see if anything comes of it. For now, that is going to do it for this weekend episode of the AI Daily Brief. Appreciate you listening or watching as always, and until next time, peace. Peace.