Nate B Jones reviews his AI predictions for 2025, identifying what surprised him most
A solo reflection by Nate B Jones on what actually worked — and what didn't — in AI during 2025.
Summary
Nate B Jones of AI News & Strategy Daily reviews the AI landscape of 2025, identifying nine developments that surprised him — mostly positively. He argues that two foundational unlocks drove much of the year's progress: LLMs gaining the ability to use code as a tool, and the solving of high-quality image generation. He contends that the "messy middle" layer of AI infrastructure proved far more valuable than many predicted, that agents work best when properly scoped rather than treated as magic buttons, and that creative problem-solving instincts became the most in-demand skill in the AI labor market faster than expected. He closes by noting a shift among business leaders away from cost-cutting as the primary AI use case, toward quality improvement and scaling customer value.
Key Takeaways
FULL TRANSCRIPT
The Two Biggest Unlocks of 2025: Code and Images
Nate B Jones: 2025 didn't deliver the science fiction version of AI that gets lots of clicks, but it exceeded my expectations in ways that matter more. It clarified where value is actually coming from in the AI revolution, and it made the gaps that we still have visible in a way that I think is useful.
The number one thing I want to call out as we look back at 2025 is that almost all of us underestimated how powerful it is when you allow an LLM to use code as a tool. That turns out to be an absolutely massive unlock. It's at the heart of a lot of our agentic workflows these days, and it is on the verge in 2026 of being one of the biggest places where non-technical folks can lean in and start to figure out how agents can work for them in ways that are far removed from writing code.
That core unlock — that an LLM can work with code, and therefore it can work with any part of the computer — was something you could see the model makers had in their heads at the beginning of 2025. They did talk about it, but it wasn't realized. It was very much a "someday we will see this come true." And then over the course of the year we got Claude Code, Model Context Protocol started to dominate, we started to get skills, we started to get Codex, and you gradually saw these pieces come together. You saw the development of Cursor, and so on. What you start to see is that these pieces allow everyone — not just technical users — to start to use the tool, because now plain English allows you to talk with your computer. Plain English allows you to manipulate the files on your computer any way you want. Absolutely massive unlock. And I think it is hard to correctly estimate going in how big that turned out to be.
The second one I think turned out to be absolutely massive is images. We lived in a world for most of 2025 where images were getting better, but text remained the most accurate way to work with LLMs. In that world, code is just a subset of text — accurate code is just another language that the LLM has to learn. But images are how humans process information quickly. And when we finally got to the point where images were solved — where you could do detailed infographics, detailed text-in-image without it looking weird, full maps, layouts, slides — I don't think we realized how big an unlock getting that right was.
And that doesn't just mean we solved PowerPoints, though we did solve PowerPoints. Think of it more broadly than that. Think of it as generative UI for gaming. Think of it as having a chance to redecorate the rooms in your house in ways you never did before. There are all kinds of personal applications people are building around fashion that come with this. Basically, getting images right enables us to realize a vision of graphical user interfaces that has always been out of reach — the idea that the graphical user interface isn't locked to your screen and locked to what the developer says, that it becomes something that evolves with you. Maybe you end up with it as a wearable. Maybe it's a combination of phone and laptop. Maybe it's generative on the side, the way it is for Comet. Those are all variants of the idea that with the right graphics solution from an AI perspective, that whole surface becomes continuous, and you can evolve the surface of digital engagement with where an individual is at and what they're looking for in that moment.
We're not all the way there yet. I'm not here to tell you that the future of the web is everybody having a generative interface that changes for everything they need. There is value in habit. There's value in steadiness for common use cases. We don't need to reinvent the wheel where it's incredibly obvious. But solving images is one of those things we probably underestimated going in, and looking at it from the other side a month or two in, it's absolutely massive.
You Don't Need AI Developers — You Need Systems Designers
One of the things that surprised me is how far you can get without AI developers. We were told at the beginning of the year that AI developers are everything — that you need to have a developer who knows AI. I actually think what you need is someone who can design systems. I have watched individuals outexecute entire development teams because they treated engineering as a workflow they could design, and they didn't worship at the altar of a particular model. They were okay building low-tech things like templates, validators, and retries. They iterated really quickly and aggressively. They didn't confuse "agentic" with "good." And before long, those folks were understanding the new principles for evolving agentic systems and becoming more and more valuable.
I think we are going to have to throw away the idea that there are technical and non-technical people. A more accurate description is that everyone picks up the degree of technical skills they need to solve the problems they're interested in. Increasingly, the question is going to be: are you curious about the problems that are relevant in your domain? And are you willing to dive in and pick up the AI skills you need to solve those problems — including technical skills — because those are increasingly approachable? You can get a scheduled task with a learning session composed just for you in ChatGPT every morning if you want. I do — I get nice little coding reviews every morning. You can get it for whatever you want.
Verification Loops as a Supercharger for Agentic Systems
Another thing that positively surprised me about 2025: verification loops in agentic systems turn out to be incredibly powerful. The idea that you can measure correctness in different dimensions turns out to have incredibly wide-ranging implications for good system design. That's not super surprising if you know how software is designed, but hooking that up to an agent that iterates is like hooking a jet engine to an airplane. It's amazing how fast you can go when you put an agent against a verification loop that is hard to game and say, "Go get it done."
I think we figured that out as a community partway through the year and started really practicing it, and that has supercharged our progress. The nice thing is that's one of those primitives we can really build on heading into 2026. Getting more into verification loops is something I think we'll see from more and more teams. We're going to start to see some consistency around those — things like accessibility, where you just want a standard set of evaluation verification loops across the industry and people just need to get their agents to pass them when they're building software. It should not be something you have to reinvent every time. So I think we're going to start to see a really interesting ecosystem build up around verifications.
The Messy Middle Turned Out to Be the Entire Game
Another one: the messy middle turned out to be the entire game. Everyone wanted to talk about the front end, and there was a lot of talk during the middle of the year about super model makers or hyperscalers owning the entire stack — ChatGPT owning the stack, Claude Code launching and owning the entire stack, whether Cursor was game over, and so on. It turns out there is so much value in transforming messy inputs into structured representations, in routing intent, in orchestrating calls, in handling exceptions, in providing useful user interfaces for specific things, that the middle layer is still feeling underbuilt relative to how much value we can unlock.
I think most of us got that one wrong. And I think that was one of the pleasant upsides of the year — the messy middle, yes, maybe it's vulnerable, maybe you worry about the launch of a particular hyperscaler's product, but there is so much value in taking raw AI model outputs and getting them into a particular domain that we are really underbuilt. Probably the most prominent example is Cursor, which everyone refers to all the time, but there's a whole host of other startups in the space — not just in coding, and in non-tech spaces too — that are building aggressively into the middle because we've realized we don't have to be afraid of it. We've realized that the model makers are essentially forming a very competitive substrate of intelligence that we can build over the top of to deliver outputs to users that are much more valuable than they're really going to be able to get from models alone.
Properly Scoped Agents Deliver Real Results
Another positive reflection from the year: I think we are realizing how much value there is in effectively scoping our workflows. Agents were oversold — that was something a lot of people were disappointed by because they were sold as magic buttons. But the flip side is that when you put an agent in a good workflow, that's a really pleasant surprise, because instead of promising it can do everything, it turns out that when you use it as a tool, it can do a tremendous amount reliably and you can start to really move volume over to it. The pleasant surprise is how much you can accomplish when you properly harness your agents, and how big companies are leaning in and actually getting volume done on that basis.
AI Slop Is a Symptom of Poor System Design
We had a lot of conversation around AI slop this year, but one of the things I learned is that AI slop is a symptom of unconstrained and unmanaged artificial intelligence. Companies that get into marketing and producing content at scale with AI — if you build the right systems — can produce really compelling, very performant ad flows, very performant email marketing, very performant content marketing that outperforms what humans can do.
The hopeful thing there is that we don't have to surrender to and live in a world of non-performant slop. We can actually construct these systems so that they're beautiful, so they sound good, so that people want to click on them. I'm fully aware that right now, if you announce that your ad is AI, there's generally a backlash. I think we're just going to get to the point where we don't announce it anymore and we just do the ad that feels right — however we do it, whether that's manually, manually plus AI, or fully AI — and we just get it done and move on. I think that's going to be the case with a lot of content.
The key measure we should hold is: is the content useful? Is it genuinely helpful? Is it information-dense? Is it something I can come back to again and again and find something to dig my teeth into? What we're discovering is that AI can actually be really helpful on that, because it can ground you with research, it can increasingly do fact-checking, it can help you think through the structure of a piece, it can help you generate ads now that we've solved the images piece. That gives me hope, because I don't want to live in a world with slop and I don't think most people do. I think that's going to end up being a phase we get through, because there is a lot of selection pressure for better content.
The Labor Market Selected for Creative Problem-Solvers Faster Than Expected
Another thing that surprised me positively in 2025 is how quickly the market selected for people with very strong creative problem-solving instincts. I think we saw a quicker response from the labor market than I anticipated — not on the dramatic headline stuff of firing and all of that. There's actually still not a ton of evidence that AI is driving overall job market declines as much as there are newspaper headlines about it everywhere. But there is a lot of anecdotal evidence around the degree to which AI is a creative or liberal arts endeavor. And I think that signal has swung really sharply.
That's been really positive to see — technical people who wanted to express their creative side, and creative people who never felt like they could be technical, finally having a chance to get the AI they're looking for, a chance to build, a chance to share their talents, and a chance to stretch their wings in ways they couldn't before. One of the great opportunities of 2026 is for people who want to grow the edges of their domain expertise and get smarter and do more. The world is your oyster. We've never been able to lean in more that way, and I think that's been really fun.
The Shift from Cost-Cutting to Quality Lift
The last one I want to call out is that we did start to see a shift from a lot of the cost-cutting mentality to quality lift. It's not universal — there are absolutely still people who see AI primarily as cost-cutting. But more and more, as I sit and talk with leaders, they're recognizing that after the first wave of vendor purchases, getting AI installed quickly, and thinking of AI as a magic button, they still need their people. Their people can't go anywhere because they still need them to deliver the kind of value that only people can deliver in a customer-facing organization.
So what they want is a conversation about quality. How can we level up the quality of the experience we provide to customers in ways that were unimaginable before AI? How can we lever up the volume of customers served? How can we make the price more competitive because we're able to scale the unit economics of the business? Those are so much more interesting and compelling questions than simply asking whether we can cut costs and dump AI in. There will still be folks with that brutalist mindset. But increasingly, people are starting to recognize how powerful these systems are, and they're starting to recognize that the firms that win are firms that regard their people and their people's attention as a precious asset — and that are designing AI systems around them in ways that allow people to put their expertise to work where it matters most.
So my question for you is: what did I not mention? What exceeded your expectations in 2025?