A (0:00)

OpenAI has made a major change for developers when they're developing AI agents, how they're going to be essentially making these things more effective. So today on the podcast, I want to break down everything happening with this new update. This recently. They made a tweet about this a couple days ago over on Twitter. So some really, really interesting new developments that I think are going to have some big impacts on when we're actually getting these autonomous agents that everyone is talking about. So let's dive into it. Before we get into it, I wanted to say today's episode is sponsored by my very own podcasting course. If you are looking to set yourself apart in your industry, stand out, and kind of elevate your personal brand or your business, I believe that a podcast is the number one way to do it. Obviously, I'm biased because you're listening to this on a podcast right now, but I will say that this podcast I started is the best. I think it's probably the best financial and time commitment or investment I've ever made into my own businesses and my personal brand. And I think that everyone should start a podcast. Again, I know I'm biased on this, but I truly believe, you know, I've been able to get over 4 million downloads on my podcasts. I've gotten thousands of customers for my software products and have been able to raise hundreds of thousands of dollars, all thanks to my podcast. So I highly recommend anyone do it. If you're interested in getting started on a podcast, I recommend you take my podcast course. I've been podcasting for five years and my goal really was to make something that you can essentially skip all the mistakes. And there is plenty that I made over the last five years and get straight into how to create a really successful podcast or grow your existing podcast. How I was able to get 4 million downloads, how I was able to create a podcast that would, you know, get people in the door and help find customers. So if this is something you're interested in, I highly recommend checking out this new podcast course I've launched. Let's get into the episode today. So the big news here is that OpenAI essentially is making it easier for developers to improve. How AI assistants, or it's kind of, they're calling them assistants. It's kind of like the first step, they say to AI agents, these autonomous AIs that can go around and do things without us. So they're making easier to implement how for developers to implement how these AI assistants find and use information from files they've made some updates to their API. This is going to help assistants give better answers. The update that they've just done essentially is letting developers have more control over the results that the AI selects, which is going to make it more accurate and also more useful. The responses that it, that it gives. This is something that a lot of people are very excited about. I want to break down some of the implications of this, but first of all, some of the responses. So Simon Wilson over on X, he said this looks like a big deal. I've mostly, I've been mostly ignoring OpenAI's reg offering so far because without details of how it works, chunky strategy, relevance mechanisms, et cetera, it wouldn't make informed decisions about how to effectively build with it. That seemed fixed with these changes. So the actual tweet that they put out, it's over on the Open, it's at OpenAI devs, it's the OpenAI developers X handle. It says, we just rolled out technical controls for file search in the assistance API. To help improve the relevance of the assistance responses, you can now inspect the search results returned by the tools and configure their rankings. So what does this actually mean for, you know, the tools that we are all going to get and we're all going to be able to, to be actually able to use? I think really the big thing here is that these are essentially designed to help these AI assistants get better results. That's, that's the bottom line. Developers, they're now able to essentially adjust the, these AI agents, how they select information, how that they're using to generate responses. They're able to determine how that's done, especially with files. So right, like you imagine like files on your computer, these AI agents going to the files on your computer and grabbing something, essentially it's, it's easier to actually do that. And this is really interesting because right up until now we've done a lot of these AI agents or kind of bots or assistants or whatever. They're going to websites, right? They're searching the Internet, they're finding some information for you, they're returning it, they're doing things. Is we're getting to a really interesting point where all of a sudden these agents are going to start actually doing actions on our device. They're grabbing files, they're moving things around. It's going to be a lot more in depth than just web based. And I think it's interesting because we're seeing this, sure from like a computer perspective, right. Like this agent could go to your computer, could Open up a file, grab a PDF out of it, share it with some other, you know, some other location that's interesting or like you can imagine if there's some sort of agent that helped with video editing, it could like know the right folder to go to get video clips for something. So like that's a very interesting concept, but I think we're going to be seeing this also in a big way very soon with Apple's Apple intelligence, where it's going to be grabbing files and doing things on your actual phone. So we're seeing this on the desktop, we're seeing this on phone, we've already been seeing this like with web for a while. All of these things are going to start getting tied together and these AI agents are going to be like living and moving around and doing things on our actual devices, which I just find so, so fascinating. So essentially the new feature, it's just going to let developers inspect file search results and modify the rankers behavior to make sure that there's relevant results are being prioritized. This like whole assistance API thing, this was launched back in November of 2023. So last year OpenAI said that this was quote, a small step towards creating a fully autonomous AI agents. So they're acknowledging like this really is kind of the, this API is like really how we're going to get there. And they just said like this is a small step, we're going to make a bunch of improvements. So this is one of the big improvements that they have rolled out. So the API essentially is going to allow developers to use OpenAI's models with specific instructions and integrations and they're going to be able to integrate other OpenAI ecosystem tools as well. Right. So you can imagine you're going to be able to grab OpenAI's voice model, start sticking that in, probably eventually OpenAI's video generation model, Sora start attaching that to this. So everything's going to be able to branch out of this kind of assistance API where it's going to integrate with all of OpenAI's AI tools. So it's going to be able to generate images, video, text, audio, talk with you, do the visual, to be able to see stuff and get context from that, a lot of really interesting things and it all gets tied together in this API. So really what's another interesting thing about this is that this whole assistance feature of OpenAI was created and they created this whole API so that it can actually interact with other agents. Right? So you could have one agent, you know, maybe in Charge of your. And, like, sure, I'm spe. I'm speculating here, right? So, like, don't, don't assume this isn't immediately available. But you'd have one agent that is your lawyer and one agent that is your accountant. More likely, you could have agents that are like, helping your lawyer, helping your accountant do things. Right? Like, maybe it's going to. You have a lawyer and they're able to get all, like, legal documents from it or something. In any case, you have these different agents actually working together. And so this API is helping to tie all your different agents together so they can all work together, which I think is so, so fascinating. There's a ton of really interesting new tech that's also come out as far as, like, these agents go and how this work. I just recently kind of covered one of the really interesting companies that's creating, like, essentially bank accounts for these agents. So obviously you need a human to own a bank account, but they're helping to make like, a programmatic way of. You could have different agents and you can give them different budgets for different projects they're working on. You give them actual bank accounts or credit cards. It's not just coming out of like a big fund for the whole company or department and they have to, you know, say everything that they've used their expenses on. It's like, very transparent. You're able to see when, what, why, and which agent. Right. So you might have like, seven, like, accounting agents or seven, like, customer support agents, and some of them might be spending money on different things. You want to know who's spending money. It assigns. It assigns values and it assigns ID numbers to every agent. And so you can know, like, who's spending money on what and what they're doing. There's all these really fascinating things that are happening with agents right now, and I think this is a big area to watch those. OpenAI keeps making these. Sometimes it might seem like small changes to this assistance program. This is where you want to watch because this is what the agents are. A lot of these agents are going to be coming out of opening, obviously, is the biggest, and developers are going to be using what. What comes down the line from this. So I think initially OpenAI described this API as enabling assistance that still require guidance. So that's why they're calling them assistants and not actually agents, because these are not things that are fully autonomous. But it is a step towards kind of having these more independent AI applications, I think. So so far, like I mentioned, developers have responded really positively. To the update. Some are saying like they're, you know, they're thrilled. This is going to help them to essentially better tune their assistance. McKay Wrigley said. One of my biggest wants, Nick Dub said this is very intriguing. Will GPTs get in app options to control this too would be very nice to modify. Lot of people seem to be quite excited about, you know, a lot of the, a lot of the features that are, that are being developed. So right now AI agents I believe are really going to take over a lot of tasks that have minimal input. So you can think of things like booking flights or automating data processing for big, large companies. Companies like Google Salesforce, they're already working on different agent platforms. Google has Oscar, which is becoming open source, um, and then Salesforce has been, you know, releasing a bunch of enterprise specific agents lately. So a lot of these big companies are working on this. You know, opening is not the only one. AI agents right now, I think really they're in their early stages. There's a lot of room for improvement, specifically in kind of accuracy, which is where we're hoping that this new feature, this, you know, there's file search controls that when I just released to their API, hopefully that comes into play to help make that better. Benchmark tests on these different AI agents currently, I don't think have like a lot of comprehensive, comprehensive metrics to really evaluate how well these agents are. Right. We have a lot of like benchmarking platforms for just determining how good like ChatGPT versus Claude versus Gemini versus, you know, llama are. Right. So we have these like benchmark kind of arenas or whatever where you can test the models. But really there's not a very solid version of that for agents. Right. So it's hard to know what's going to be the best. But this is no doubt going to be a very big area and something that's going to be very popular in the future. People testing how well agents do at different tasks. So overall, very, very excited in every update. I think that OpenAI does is definitely bringing us closer to having your own agent. Or maybe you'll have an agent on your team at work that is in charge of specific tasks, running around, doing things. It's going to be a fascinating future. So I'll keep you up to date on everything that happens in this case. Thanks so much for tuning in to the podcast today. Hope you learned something new about AI agents and what is going on in the field. If you are interested in starting your own podcast, there's a link in my show notes. It will be the best podcast course you ever take in your entire life. I can promise you that. And I hope that you all have an amazing rest of your day.