OpenAI Claims 2026 Is The Year AI Agents Transform Everything We Know

2026 is going to be the year of the agents. We've been through chat. We're seeing early signs of it. Yes, there's naden. Yes, there's make. Yes, we've got some of the early tooling of multi-agentic flows of totally auton autonomous flows. But there's more to the story. Let's jump into it. Don't listen to me. Listen to Sam Alman and Brett Taylor, the chair of OpenAI. had a great talk at Sequoa talking all about his new company where he's saying, "Look, we used to have websites. That was the thing that we needed. Now, everyone's gonna have an agent, both personally and professionally, big businesses." His company does the B2B sort of agent space. Now, there's lots of agents underneath that one agent, but he's also a master at sales. So it's a lot easier to just say you have one of them when right now it's actually going to be dozens that are lattering up to one. So we have that going on. We also have another really great talk like I said with Sam Alman where in his talk he's basically look this year we're still figuring out early signs of agents. We're cobbling them together. 2026 is going to be the year where agents are actually delivering tons and tons of business value. And then going into 2027 is when we're going to see the physical manifestation or the fruits of our labors from AI coming into the physical world. That's through robotics. So kind of bananas. Highly recommend watching both of these talks. I actually recorded a video for each one of these, but then something was wrong with my audio of course. So beyond that last point on this talk if you don't watch it is him basically saying that this shifts the entire pricing model of SAS of software where we went from on premises or sorry we went from selling software in a box I don't know if if you're like not 30 years old like I'm 32 and I don't even really remember that but we sold it in a box and we went from box to on premises to the cloud and then from the cloud to a software as a service where it's basically in order to cover R&D we needed to charge monthly but they got updates and now it's actually going to be paid by the job which is really cool and it aligns the builder of the software with the company which is great. So all that said let's look into some some of the tech that we as builders can be leveraging. So today is Wednesday, May 14th. It's been about a week since I've made a daily video. If you're new to this channel, I'm Parker Rex. I led a tech for a startup that sold for 23 million bucks. And I've been using AI every single day basically since OpenAI launched all of their SDK stuff. And yeah, the 3.5. So it's been a few years now. And I talk about it. So, first of all, there's a great Discord that I found that just tracks the different models as they come out. So, I don't have to be a gatekeeper on technology or where I get these things. You can check it for yourself. This is our Discord and the community called Vibe with AI. We're coming up on 100 members. Really excited about that. It is not for brand new people that don't want to put in the effort. Who it is for is people that are willing to put in the effort if they are brand new or if they're enthusiasts that are trying to get better. But we discussed a lot of stuff in here. Even something yesterday that came up was cursor released something and then took it away and I covered that. It is a remote cursor protocol that is coming out soon which means you'll be able to control cursor. But let's go into this channel. It's called dev mode. And if you comment below, I'll drop the Discord link or you can just freeze frame it. Hidden crap. Well, just comment below if you want that or DM me. Just Parker Rex on Discord. But we can see an entire thread of just new model alerts as well as app updates as well as web watchers. So, I actually want to build this into our Discord. I'm like, this is sick. But yeah, so what came with that is basically Google is absolutely ripping. They've got Imagine 4, Imagine 4 Ultra coming. Speaking of agents, they're going to be releasing something next week that changes how we use Chrome. So if you're not using Chrome, you're in the minority. Majority of people use Chrome. And they're going to be bringing Gemini right into Chrome, which is going to be absolutely crazy. It'll be a lot better. I use Brave right now and this Leo thing it's a nice idea but does it work well now so anyways on the agent side we can see that the Google chief scientist Jeff Dean total gangster let's just actually try this out what are some of Jeff Dean's accomplishments still learning how to doesn't have enough. Great. See, ex case in point, it's useless. But he says, "We will have AI systems operating at the level of junior engineers within a year." I think that's a much more sober perspective than what some other CEOs are saying. On the cursor side, we know that they're putting background agents in place, which I'm very excited about. They also launched a bunch of other stuff in here. Agents in the cloud, bulk context tools coming soon or more to come. inline edits to agent flows, faster edits on long files, multiroot workspaces, better pricing, yada yada yada. They're really focusing on context and also within context, mermaid diagrams. So, like I always say on this channel, Klein leads the pack on innovation. So, I always flip between Klein and one other tool just because I want to see what's going on with Klein. And I followed a lot of their patterns without even using Klein. So, I can just put in the AI docs that's from Indie Devdan. If you don't know who that is, definitely go check out his channel after mine. But on the agent side, I'm really excited because soon I'll hopefully get access to the background agents where then I can run multiple agents at once. And I think that there'll be a lot of work done, a lot of complaints too around lack of dark mode, but I think a lot of work will be done on this right here where right now the custom mode and all this advanced optioning really not that fun to use to say the least. But I do have one with memory bank in it. And what that does is it takes client's memory bank and puts it in there. Then you can bind it to keys. But I think this will be the future is essentially we started with cursor composer going back maybe six months and that was an overlay and then they brought it somewhere else and then they brought it all into one spot and I think the next iteration of this will look something more like a dedicated agent orchestration user interface and the jobs to be done for that will be where is my agent at in the current process. As a user, I need to start a new work thread. So that's what that is, a workflow. I need to start one. I need to see where it's at. I need to be able to easily manage the different statuses of how far along they are. I need to manage the context. I need to manage the custom instructions for these things, the set of rules. So you can apply rules, the set of tools. So all these things, I think that's going to be incredible. Now, I'm really really bullish obviously on cursor, but I do need to mention after I spoke with a team member at augment last week that I've uninstalled it because it was conflicting of something in here. But I had a 30-minute call with a VP there and what I gained from them is they're all in on context brains as they call them. So they have a big group of engineers that came from Google and they're all in on not making the best co-pilot. They actually want it to be for advanced programmers that have large code bases with lots and lots of lines of code, lots of complex things where context is going to matter a ton and that's a big problem to solve. So I'm really excited about that. They also launched their remote agent. If you want to go sign up for the dashboard, you can use this link. I will put it in the description or you can just go to augment.new and sign up. And what else did they mention? Basically, they're aware that they're 18 months behind on the kind of VS Code fork vintage aka wind surf and cursor. And so they're really doubling down on context and it is a big problem like especially in new code bases or or large code bases. So really really looking forward to getting access to both the background agents from cursor as well as the new remote agent from augment. Other stuff that's going on, if you want to check out a really cool open source codebase, Firefox just moved to GitHub. So that you know now it's on GitHub. So that's cool. And what are other things? Oh, just cursor's releasing a lot of little quality of life things. So you can tell that they are cooking because OpenAI bought Wind Surf. They need to cook and little things like the way that you manage chats. If you're ever running into performance issues, just clear out your chats. It'll make it better. And then codebase indexing. So they purchased Super Maven a while ago. and Super Maven six 8 months ago was kind of the goat in the space. So I know that he's probably taking a big hand on the codebase indexing. They now do get graph file relations where you can change it. You can decide in another setting if you want to have all of the files that are mentioned in a folder included in your context. So it's basically just better control over context. If you don't have any of the features that I'm talking about, then you need to make sure you have beta enabled. So, go to beta, turn on early access. I always check this out. This was where that remote cursor protocol was, but then it went away. And there's a good article on that that I mentioned in our Discord that basically I'll just play the video real quick. In the video, I basically asked if anyone else had access to this. You could see that this was a checkbox and then there's some rumblings on the cursor forums just explaining how you could enable the RCP server and then how that would have different use cases where it's oh cursor is kind of going into a space they haven't touched before which is you could control cursor through this RCP so MCP but RCP so you would be able to either have custom ID extensions that can control your cursor you could have automated code review. You could have CI/CD, which is pretty interesting, but basically opening up cursor in a way where you can manipulate it. So, I'm trying to think if there's anything else. Yeah, it will kick off a new codebase index when you've updated to the new one. If you're watching this, you probably already have updated, but I just love to see when we make calls on this channel that are correct where I had said a while ago, hey, client has these mermaids. You're going to see this in cursor soon. So when you want to live on the bleeding edge, you go use Klein. It is always ahead. Just literally always. They also have really good prompts. So if I go to Klein and I go to the memory bank, you can see that they have all these different rules in here. So you have rules for the architecture of how they actually do it. They have rules for the continuous improvement protocol. I have not used this one. Maybe need to check it out. The one that I always am using is memory bank. As I've said, I've tried different iterations of this where I think as you scaffold a product and in my case I have a monor repo with fast API and tan stack start and superbase that as I was going I was changing the way that I was using the different AI tools. So when I knew less about the stack, I'm using less AI tools in order to force the learning to get better. But then as I'm advancing and I'm getting better at the technology because I'm total Python noob, then I'm putting on more and more AI tools. So the way I kind of scaffold this out in a real life example was I just asked a million different questions and I'm just basically spamming the ask over and over and over again. And then when I felt better about it, that's when I actually put individual memory banks into each module or each area of the application. So I had a memory bank for the database. I had a memory bank for the API and a memory bank for the web product. And that just allowed me to learn more. And then by forcing myself to just ask questions over and over again and then pair that with research, I'm paying down that that gap basically that idiot tax, whatever you want to call it, ignorance searchcharge so that I could get through it a lot quicker. Whereas if I just jumped in and prompted my way around, I wouldn't actually be competent in the codebase. And I think that you are going to pay for it either way. You're going to pay for your lack of knowledge with tokens and lots of them through just letting it make code and that's okay, but I prefer to use my tokens and get the knowledge in my head so that I'm a more well-rounded developer. You can see around corners. This is something I talk about a lot with our members is look, I'm not your guru. I'm just kind of building it in public. I'm sharing what I know. And in this case, I know there's secondary and tertiary effects to actually learning the technology so you can see around corners. You can you can basically change the way that you think as you learn more of these languages and patterns. You get better at pattern matching. You get better at systems thinking, systems planning, and your product strategy in general basically gets better. I look back to when I was a product manager at our company for it's five years of product management plus UX design at the same time. UX design for two years and I thought I was technical because I could read front-end code just basically like vanilla React. But having gone through the journey of now coding pretty much every day for three years on top of starting and stopping coding for 15 years. I compare myself now and I'm like, "Wow, I wish I started sooner." So, if you are new and you're just prompting your way around, I highly encourage you to pick up the weights that are learning. I have a community. I have a lot of people that are joining it. And I make this analogy where I think it's a lot like if you got excited to go to the gym and you saw someone that was in great shape and you were like, "Wow, if I just have the stuff that they have, if I have the prompt, in our case, in this analogy, if I have the running shoes and the protein powder, then I'm going to look like this person." But the reality is is like those are just enhancers. So in our case, AI is an enhancer. It's a multiplier. It is leverage on top of existing skills. And what's beautiful is that you can get a better hello world. Like when I first started coding when I was like literally 14 and I was like running Counterstrike servers. Shout out Counter-Strike 1.6. I got to make my first hello world with HTML and that was awesome to see it on the page, but not that awesome. Now you can go on to Bolt and you can make an awesome first marketing page, but it's almost a double-edged sword. So, because then you have to make this leap. That's it for this video. If you want to check out the community, you can see the link in the description. If you found this video helpful at all, make sure you subscribe because I make one every single day. I was gone for last week because I was at a wedding. My wedding. And I want you to like the video. Cole, did you like the video or did you think I paused? Anyways, I got to go and I'll see you in the next one.