Transcription
If you've been following AI, then you know 2025 was wild. New capabilities dropped constantly. Whether it was video, image, music, or code, every single modality in AI leveled up big time. We went from "AI can do this" to "AI can do this, this, and even this" in less than 12 months.
But 2026 to me feels different. It's less so about what's possible anymore and more so about what actually works. So this is the year of maturity, of depth over breadth, and most importantly, of mastering instead of constantly chasing the newest thing.
So in this video, I'm going to walk you through who's winning right now. And it's probably not who you think. There's a core mindset shift from the people actually getting results versus everybody else that claim to be. So, I'll give you the mental model playbook that you need to position yourself as best as possible to win.
The individuals and companies that will win in 2026 aren't going to be the loudest ones. They're not going to be the ones with the biggest announcements or shiny features. They're going to be the ones that keep shipping quietly and responsibly as everyone continues to be distracted by the latest and greatest thing. And the biggest example of this might surprise you.
So this time last year, we were asking existential questions like, "Will Google as a company cease to exist in this AI space?" After years of watching things like Bard AI come and go and having multiple Google mishaps, the question started to arise as to whether or not they would be a part of the conversation. And if you've kept tabs on the company in the last few months, they've been shipping at a velocity unseen from any other company. And what's crazy is they've been doing this at a rate where all the products are actually useful, they're affordable, they are developer friendly, and most importantly, they're becoming they are the best AI generalist company. So whether you want an amazing image model like Nano Banana or a video model like Veo, they've been able to cater all of the generalist skill sets of a generative AI model. And as of this recording, Gemini 3 Flash is the standard for search for billions of users across the world.
Now that the Gemini ecosystem has matured, they have the leverage of hyperdistribution. Google has distribution that no other AI company has. And the biggest twist is that they are self-sufficient, meaning they have their own infrastructure, their own TPUs, their own servers. They don't need to depend on companies like Nvidia to give them more compute and more chips to keep the ball rolling and keep progressing as a leader in the space. Especially when it comes to enterprises that are constantly solicited by hundreds of AI companies that each individually do one micro thing. Them being able to say, "Oh, it's already in our Google AI workspace," is a killer argument.
In other words, I see companies like OpenAI having a lot of mind share versus market share. Mind share, meaning they're probably going to really double and triple down on being the most consumer-friendly product from a back and forth chitchat standpoint. But since companies like OpenAI are not self-sufficient, meaning they always have to go and look for new training data, new revenue streams to justify their crazy valuation and raise more and more money, companies like Google can continue their dominance.
And this really connects to a much bigger pattern, which is that the winners are not sprawling. They're very focused on exactly what they're setting out to do given their capabilities and given their distribution. And if you need yet another example, look no further from Anthropic. Has Anthropic come up with an image model, a video model, an audio model? No. They focused on being the best enterprise LLM that they can, building the best APIs, being the most developer friendly, and building, in my opinion, the best agentic coding platform in the form of Cloud Code. And if you follow the channel for the past couple months, you'll know that I've really gone deep in Cloud Code. And the deeper I go, the more thoughtful features I see, and they just keep shipping over and over again, taking all the feedback from the market and reincorporating it.
In contrast to this, you had companies like Meta spraying and praying, releasing Llama 4, which is probably one of the least well-received open-source models ever for the cost that they put in. And they lost key AI personnel and had to consistently pivot to find a new market for their product. And just a few weeks ago, they acquired a very popular platform that I've covered on this channel called Manis for more than $2 billion.
So, the counterintuitive point that you could use to make 2026 as successful as possible is: just because you can do anything now doesn't mean that you should. Winners have restraint and they go deep where everyone else goes broad.
So, that's who's winning and the mindset of winning. But here's what's really interesting to me. It's not just companies that have changed their mindset. It's also how people are thinking about AI in general. 2025 had a certain energy. 2026 has a vastly different one.
From what I can see, this very time last year, you had claims saying that this 203 Node ended workflow will do everything from organize your company to answer all your emails perfectly to feeding your family to giving you a massage. You had, and unfortunately still have, fake AI influencers on LinkedIn saying, "Comment 'donkey' and I will send you this completely broken workflow that I took from the Eniden marketplace for free." This mentality of blind optimism led to probably millions of half-built projects where someone started using something like a Lovable or a Bolt and said, "Wow, look at what I built. I built an interface that I could never do because I don't know how to code." And it really just stopped there. There was no taking it from A fully to B and actually deriving value from whatever you built. So there was a lot of building for the sake of building, adding and creating a SaaS that pretty much no one will actually subscribe to because by the time you're done the SaaS, the functionality is possible to create on your own using something like Cursor or WindSurf or Cloud Code.
So this is the year where the winners will ask this question: "Will this actually work for my use case?" The good example of this are AI agents and everything that this concept of agents promised to the world. 2025 was supposed to be the year of the agent, but in my opinion and from what I can see across the board, it wasn't. It was the introduction of agentic coding. There was a lot of confusion that if you had six or eight agents in your Cloud Code last year, then you were super productive. They worked perfectly. They worked in parallel. They never overlapped and they all worked together to accomplish the same goal. In reality, though, they took and burned tons of tokens. They built overlapping code that broke each other's contributions to the codebase. And there was mass confusion over the understanding of the difference between an agent, a prompt, an MCP, and a tool, and how all of those need to work in unison.
Especially when it came to autonomous agents, where the promise was if you quote unquote hired 10, 15, 20 swarms of agents, then you could accomplish days of work in just minutes to hours, was completely falsified. If you had workflows, even with the smallest bit of nuance, you would see 95% accuracy from the first step to 90% to 80% to 60%. And the more steps, and there was actually studies around this, the more steps that were added to an agentic workflow, the lower the accuracy every single time. And especially if you're trying to use this in a production use case, meaning actual day-to-day, client-facing, customer-facing, this wouldn't make sense whatsoever. It would make way more sense that you just assume you build without agents and you bring on an agent as if you hire someone in a company and you're not a VC-funded company. You hire an agent when absolutely needed and you need a parallel use case that fits perfectly where there's aligned incentives. There's no overlapping goals.
So, I still don't think that by the end of this year we'll have perfectly running autonomous agents that all work in unison without any form of human in the loop. What I do think we'll see is an exponential increase in the ability for agents to actually do agentic things that have been promised to us for the past 12 months. Meaning, you will be able to have three or four agents go and work together and ideally be able to communicate with each other effectively and efficiently enough that they don't step on each other's toes. And we already saw a preview of this last year with Google's A2A framework. And this is pretty much a medium through which different agents can effectively and efficiently communicate with each other to increase the likelihood that they don't step on each other's toes and actually can work together to accomplish one goal.
Last year, everyone and their grandma came up with a vibe coding platform of their own that did pretty much the exact same thing. And I can tell you firsthand, my inbox kept blowing up day after day with all of these no-name companies asking to sponsor the channel with their vibe coding platform that competed with the likes of the Bolts, the Lovables, the Base 44s that pretty much did the same thing but worse.
So, just like in 2024 where you saw a mass consolidation of all of the millions of AI tools people came up with to take advantage of these new chat APIs, 2026 will probably be yet another mass consolidation event. A lot of these vibe coding tools will disappear off the face of the earth. Only a couple will survive on top of the existing platforms like Cloud Code and Anti-Gravity and Cursor, which to me are imminent competitors to these middlewares that are browser-based. And while I'm using vibe coding as an example, this is just a microcosm of the entire space. To SaaS across the board will hyper-consolidate as more people wake up to videos just like mine that I've been releasing and will continue to release over the coming weeks to months on how to build X with Cloud Code where it's not just some flimsy knockoff of the existing app, but you can actually go from an idea to skipping building an automation or a workflow, asking Cloud to build it, building a front end, actually deploying it securely, and using it for yourself without the need to actually worry about scale.
And this is where only the valuable SaaS products will survive because you'll make the decision to pay for a SaaS product because maybe they have an economies of scale that you can't accomplish. A good example is let's say you want to build a platform that does AI editing of short-form videos. If you just want to do basic editing, you can ask something like Cloud Code to put together a whole workflow where you can upload a video and it uses a bunch of Python libraries to do it for you. But if you wanted really high-quality 4K B-roll and you don't want to pay out of pocket for something like Veo, the Gemini model for video generation, then you could use an external API or an external service that has some form of economies of scale where they can pay for millions of tokens for a fraction of the cost that it would take you as the consumer.
So, you'll have the luxury of looking at all your AI subscriptions and picking and choosing the ones that you can build yourself to run locally on your own terms on your own dime or the ones that make the most sense to take off the shelf because even though you're paying $10, $20, $30 a month, at a 50,000 foot view, you're getting $300 to $500 of equivalent value. So with a maturing landscape, you as the consumer or the business owner will have the ability to make the right decisions to pick the tools that actually move the needle for you.
Okay. So, so far we focused on who's winning and why and the general mindset shift that will happen in 2026. What is the playbook for you to win given all this information? I've boiled this down to just two core tests.
Test number one is the daily driver test. And this is unbelievably simple yet very powerful. Do you use whatever you pay for in terms of AI subscription every single day? If you purchase something that you use only a couple times a month or does something like social media posting on your behalf programmatically, most likely you can now build it yourself with little to no technical skills whatsoever. So as these models increasingly get better and we get things like Claude 5 and Claude 5.5, we will literally get to the point where you can have a quick conversation with Cloud Code and it will be able to spin clone the UI, the functionality, hook up to actual automations without even needing intermediaries like an N8N anymore. So while 2025 was about capability, "Can this do X?", it will now be about reliability. "How well can it do X?" By the end of this year, you will practically be able to truly build anything, and it won't be a clickbait title on YouTube. And I'd also say that the behemoth cloud providers like AWS, Azure, and GCP will also focus on making themselves that much easier to integrate and deploy on. So you'll be able to build actual production apps that are infinitely more secure than the paper-thin flimsy Superbase apps on these browser-based builders that you would have seen a year ago.
The second test is the build what matters test. The same way learning can be an addiction, building can be an addiction. And I'm kind of exhibit A. I'm the epicenter of this. All I've been doing for the past couple months is just building and shipping silently. Some things for myself, some things for the community, and some things just to see if I could. Now, as an AI educator, this kind of makes sense for me, and it's not wasted effort. But I see tons of people get into the habit of finally going from "I don't like terminals" to, "Oh my gosh, I love Cloud Code or I love Cursor. I want to build everything." And you go down this rabbit hole where you build and build and build, but accomplish actually nothing. You don't build it good enough that it's finally ready for production, but you've built it to a point where you can tell your friend that you built it, but there's no actual ROI. Whether you're looking to make money, whether you're looking to sell it and make money from an external client, or whether you're trying to use it in your business or in your job.
So before you build this year, especially after you watch every video you can on these builders like Cloud Code, you want to ask the following: Does this solve a weekly and daily problem? Will I actually use it or will I get so excited that I can tell someone that I built it? What's the minimum version of this that can deliver value? Not beautiful UI, just basic stuff. Not too long ago, I uploaded a video where I walked through all these different tools on the web I tried to automate to avoid giving my credit card and my information to, like cropping images, compressing files, making videos smaller, and it was a very ugly UI. It just worked really well and it does exactly what I'm looking for it to do reliably, consistently, and without having to reinvent the wheel. And the last thing you want to ask is, am I building this for the sake of building this or am I building it out of purely curiosity? We all have graveyards of half-finished projects, including me. The discipline for this year is: build one thing, make it work perfectly, prove that it actually derives value and generates value, then move on to the next thing.
2025 was about chase everything. Pretty much FOMO. Learn more, chase more constantly. 2026 should be about mastering less, going deeper, and pushing your boundaries as much as possible, especially if you're not technical.
So, the TLDDR of this video is to focus on the reasons why the people that are winning are winning. Focusing on the fact that there will be a mass consolidation event in the market that you should be aware of and not be surprised if you see your favorite sub-products merging together or getting acquired. And lastly, how to position yourself to break limiting beliefs that you can build something because this will be the golden year where anyone can truly build anything with little to no technical expertise. It will just be a matter of investing the time to do so.
So do your best to ignore the AGI debates because AGI doesn't matter. What matters is what you can do practically with no fluff. I would also say ignore the benchmark wars because a lot of companies can game benchmarks. Focus on what actually works. And lastly, really be aware of agent hype. If someone says this can run autonomously for 15 hours, ask the questions: What is it doing? And how often do those 15 hours actually culminate in the perfect solution?
The TLDDR of the TLDDR is to stop chasing and start mastering.
And with that, if you found this video helpful to get you ready for 2026, please leave a comment. It helps the video and it helps the channel. And if you want to start 2026 in the best way in a community of hundreds of business owners and entrepreneurs pushing themselves to be disciplined and get to the next level, then check out the first link in the description below for my early AI adopters community. It is literally a highway to get access to exclusive Cloud Code content, tons of tutorials you'll never see, and more importantly, access to coaches, myself, my co-founder, Taha, to truly take you to the next level. I'll see you in the next one.