At the end of every year I like to think ahead to what the next could bring in tech. This time I wrote some of it down. Ideally I have some insight into what my profession will bring over the next 12 months, but I may also end up being completely wrong. Either way, it’s a fun exercise. I will also offer a metric(s) for gauging in a year whether each prediction was successful or not.
React will remain king of the castle
For a while everyone in frontend had to suffer the same joke: there’s a new framework you have to learn every week, how stupid, lol! I still see it pretty often. In reality, frontend web development has been relatively stable for years. React, Vue, and Angular have been the three most popular frameworks (libraries, whatever) for a long time. I’ve been getting paid to write React code for close to 10 years and in that time the only fundamental shift that I’ve had to embrace has been the move from class components to hooks. The introduction of RSCs and the focus on the server in recent releases is potentially another seismic change, but the truth is that fully client-rendered SPAs work just as well as they did 10+ years ago (better, actually). I guess you could consider the move from Webpack to Vite a pretty big shift.
Which leads to the prediction: React is not going anywhere. Data on this topic are sparse and hard to gauge; the 2025 Stack Overflow Survey might be the best we’ve got. Notice how React, especially when combined with Next.js, has a bigger score than Angular and Vue combined (both of which, frighteningly, are lower than jQuery). This is not a perfect survey question; actually it’s a pretty bad one, and likely doesn’t reflect the reality of the web which is that jQuery is still the biggest JavaScript library, or that WordPress remains insanely popular (way more than its 12.4% score). Nevertheless: React is popular, and I see no reason that React will not remain popular in 2026.
LLMs will exacerbate this. There is an enormous amount of training data for React, and React will inevitably always pop up as the first recommended choice for asking ChatGPT et al. “what should I use to build a web app in 2026?” Whether or not LLMs are good at React is a different question altogether, but that doesn’t matter at all. Expect to see more React + Tailwind + ShadCN apps, it’s become the de facto standard for vibe coders and otherwise.
Please note that I am not saying React is the best choice (hot take: there is no best choice) for building webapps in 2026. I am not interested here in litigating that topic. I will say that other tools like SolidJS, Remix 3, Svelte, and many more (hey maybe those people making the jokes had a point…) likely offer superior aspects to React, whether in dx or performance or less footguns. But: it doesn’t matter! React will remain the most popular JavaScript tool for building frontends in 2026.
Metrics for success: if the 2026 Stack Overflow survey still sees React as the most popular frontend tool. If LLMs continue to default to React for building webapps.
The true cost of LLMs-as-a-service will become more apparent
Phrase this another way: the big LLMs are going to get more expensive. I pay $20/month for Claude Pro. I have no idea how much this actually costs Anthropic, and no one outside of Anthropic does, but I think it’s reasonable to say that it’s more than $20.
Like Uber and Doordash before them, LLM services are currently heavily subsidized by gobs of venture capital investment, and whatever you call the circular ecosystem of Microsoft paying Nvidia paying OpenAI paying Nvidia paying Microsoft that now makes up most of the US stock market. Supposedly 58% of venture capital funding in 2025 went to AI startups. If you want to read more on the economics of generative AI, with a particularly dark and pessimistic tone, Ed Zitron has you covered. I don’t agree with everything Ed says, but he’s the only guy out there really writing about how strange all of this is.
What does this mean for you and me, honest code wranglers who now write and will continue to write our React UI code with help from our new BFF Claude? It means we’ll start paying for it, sooner or later. We have all been through this before. It’s a proven model for tech companies: capture as much market share as you can while you’re flush with VC cash, and figure out some nebulous path to revenue positive along the way. I believe this is called the “land grab” stage.
There are offramps that could change the calculus here. Mostly, ads. Why is Google Search free to anyone with an internet connection (who doesn’t live somewhere it’s blocked)? Ads! Google and Meta and others made, and continue to make, the bulk of their revenue from advertising. We’ve been hearing rumors for months that ads will be coming to ChatGPT, with some users saying they’ve already seen them, and OpenAI denying that. If OpenAI or Anthropic can generate enough ad revenue, then perhaps they won’t have to raise subscription prices. But really, isn’t that just raising prices in a different form? We all know this by how shitty streaming services have gotten.
Metrics for success: Claude Pro, defined as having similar usage limits with access to frontier consumer models like I have today, will cost more than $20 at the end of 2026. Or, the $20 package remains, but it’s worse (lower usage limits, less model access) than what you get today. A major LLM-as-a-service provider will openly integrate ads.
The consumer electronics industry will increasingly turn to secondary sources
If you pay attention to PC hardware news then you know that RAM is the latest victim of the AI explosion. Data centers need RAM, and selling RAM B2B to AI companies is more profitable than B2C to the egirl-next-door who just wants to build a sick PC for streaming on Twitch. So says Micron; oh, for SSDs too.
Like GPUs before, this means consumer prices for RAM and other PC components are going stratospheric. This will not improve in 2026; more data centers are planned, and hardware obsolescence is going to hit whatever it is that happens in those data centers hard.
With supply restricted and demand presumably not decreasing, consumers will look for alternative sources to find the parts they need for their build. And where else, but the secondary electronic components market, can they look? This is a very loosely defined term, but encompasses anything that’s not buying direct from OEM or companies that license from OEM.
I don’t say this just because I happen to work for a company that deals in secondary electronic components, and I promise I have no insider information either. I put this one on my list because it seems to be the logical alternative. People who want to build PCs and other electronics are smart, and they’re not going to let a little market pressure stop them. They’ll find other ways, other sources. We will see new marketplaces open up with a focus on connecting consumers with parts. The PC parts market will start to resemble the used car market, where used isn’t necessarily a dirty word.
Other players in spaces related to this will offer different solutions. PC parts are to some extent fungible; if you just want to play video games, then maybe you’re satisfied with a gaming console and will give up on trying to score cheap RAM. The Steam Machine could become the defacto standard for gaming PCs and offer consumers a fantastic solution to hardware sourcing problems (I think this is poised to happen for other reasons, like Valve having a proven track record of cheap, high-quality hardware with the software to support it, and Microsoft intent on turning Windows into a horrible platform for gaming, and anything else too).
There are a lot of problems in this space to figure out. Counterfeit, liability, logistics, it’s sticky. It’s a prime business opportunity for someone to seize the RAM shortage and offer a solution that isn’t just becoming a RAM scalper.
Metrics for success: a marketplace that can be considered the “Newegg for second-hand PC parts” pops up and is successful. Maybe something like this already exists. If so, then we’ll say that more platforms like this will pop up, and be successful.
CSS will keep getting better
2025 was a fantastic year for CSS. We got container queries, subgrid, relative color, accordion improvements…and a lot more. If you only write Tailwind all day, or still use SCSS, do yourself a favor and check out what vanilla CSS can do now, it’s pretty mindblowing (I love both Tailwind and SCSS, no shade).
The upshot for developers is that much of what used to require JavaScript is now natively supported in CSS and HTML. Accordions are the classic example: no need to track open/closed state in JS, now you have that in markup and can style it with CSS.
This is kind of a cop-out of a prediction. There are many new cool features that are already supported in some browser environments, so this is less a prediction and more “wow I’m excited about how much better CSS keeps getting.”
So, for an actual prediction, how about: usage of vanilla CSS will increase relative to a decrease in preprocessors like SCSS as well as utility-first approaches like Tailwind (and CSS-in-JS will continue to fall out of favor). If you have the opportunity at your job or side project to reevaluate your styling system, 2026 will be a great time to do so. You might be surprised how much you can replace with just CSS and HTML.
Metrics for success: if() CSS function and fully customizable select elements become widely supported in browsers, to the point where most developers don’t think twice about using them for browser support reasons.
The tech job market will not significantly improve
This is a hard one because I don’t want it to be true. I want everyone who is interested in programming and building software to be able to pursue a career in it, to be able to start as a junior developer and over time transform themselves into a badass senior who can write fullstack applications for fun on a Saturday afternoon.
Unfortunately reality has diverged from what I would like to happen, not for the first time, and I don’t see that changing much in 2026. The reasons are what you would expect: AI, or at least AI as an excuse for downsizing/hiring freezes; meta conditions around tariffs and business uncertainty; outsourcing/offshoring; venture capital being locked up in generative AI companies (remember when those guys used to pride themselves on taking risks? lol); etc. Note that I am referring to the United States market for software developers specifically, that’s what I have experience in.
There is an optimistic streak among developers that market conditions will recover to where they were a few years ago and, more specifically, that there will be a glut of business who made poor decisions related to AI, like laying off your development team because one PM with Claude can do it all now, that experience catastrophic failure and find themselves scrambling to hire experienced developers who can step in and fix the mess. And we’ll graciously say yes, and command premium rates to do it with a smug look on our faces.
I share some of that optimism; I don’t think software development is a dead industry or anything like that, and I do think AI tools are vastly overhyped and are largely incapable of building high-quality maintainable software (at least autonomously). But it turns out the market is not rational, money is fake, and that software quality often doesn’t matter. We need to temper our expectations around those understandings.
“Software developer” has for a long time in the United States been an incredibly privileged position. We get paid a lot, work pretty chill jobs on the whole, and most importantly get to do something for work that we enjoy. What’s happening is that some of that is leveling out; salaries are decreasing even as inflation rises, and RTO means now we’re back stuck in traffic for an hour a day instead of taking our daily home gym break or whatever it was you did when you were fully remote. It isn’t only AI that’s causing this, it’s just one factor among several.
Like my other predictions, there are plenty of wildcards that could change this. The new chair of the Federal Reserve could strongarm extremely low interest rates like they were in 2020 (remember 0% - 0.25% rates? crazy!) which could spur VCs to throw money around again and necessitate a mass hiring of bootcamp grads who learned React in 3 months (and cause all sorts of other effects related to inflation etc, but I digress). Prediction #2 about my Claude subscription going up in price could mean it’s suddenly not actually that much cheaper at all to try to replace humans with LLMs. Regulation of AI and outsourcing and H-1Bs could help foster a stronger domestic job market. A lot could happen!
Metrics for success: the general sentiment among developers online remains “shit sucks.” This is an entirely qualitative metric. We could try pointing towards hard numbers put out by ADP or the government or whomever, but I don’t trust those - much better to read a couple Reddit and Bluesky threads and come away with a judgement about how the tech job market is doing.
Bonus prediction: on-prem LLMs will rise in popularity
I have this vision of a small software company that runs all of their AI tooling in a closet on a few thousand dollars of hardware they bought (on the secondary market, of course). No Claude, no ChatGPT, just Qwen on some metal. Homegrown. Accessible on their LAN and VPN and nowhere else. The big players will try their damndest to make this reality impossible. Will they succeed?