The state of Software Engineering in 2026

A year and a half ago I was writing everything by hand. Every function, every test, every migration. That’s how I’d worked since 2007 and it never occurred to me it would change…

Now I orchestrate almost all of it with AI, and the results are better and faster than anything I produced before. In my private work, me and one colleague are building a product for a client faster than any agency could at a fraction of the cost. They’re happy, we’re happy, and the speed still surprises me some days.

Software engineering as it existed is over... The job hasn’t disappeared, but it’s become something else. Less typing, more deciding, more reading, more reviewing... Less “can I build this” and more “should this exist, is it right, and am I willing to put my name to it”. Owning outcomes is what’s left…

The avalanche

When producing code costs almost nothing, you produce a lot of it. Features that would have taken a week land in an afternoon. Side projects that would have died as notes in a drawer suddenly exist. That feels incredible until you realise all of it still has to be understood, maintained, secured and answered for when it breaks.

The bottleneck has moved, It used to be writing, and now it’s verification. Reading what came out, understanding it, checking it does what you think it does, spotting the bit that looks plausible and is quietly wrong. AI is very good at plausible and that is exactly what makes it dangerous if nobody is paying attention.

Which is why I’ve had to be honest with myself about the people slowing things down. In larger organisations I see code reviews dragging because some engineers still heavily distrust AI. It frustrates me, but a chunk of that caution is justified. Some of those reviewers are catching real problems, and I’ve spent enough time thinking about AI-generated PR quality to know the failure modes are real, I hit them all the time.

The organisations getting this right aren’t the ones blindly merging whatever comes out. They’re the ones that have fully embraced it and still keep proper eyes on it. Speed without verification is just producing problems faster.

Horses and cars

Coding by hand is going the way of horse riding. People will still do it because they enjoy it, and there’s nothing wrong with that. But if you need to get from A to B, you take the car.

That analogy cuts both ways though, when cars arrived, nobody needed to be a better rider. They needed to know where they were going and how not to crash.

Right now, the human advantage is taste and accountability, knowing what’s worth building. Knowing when the output is wrong even though it compiles and the tests pass. Being the person who answers for it, that’s what actually lets two people outpace an agency today.

I just don’t think it’s a sturdy moat unfortunately.. Taste is pattern recognition, and pattern recognition is precisely what these systems keep getting better at. Accountability is partly a legal and social construct, and those can also shift, and if AI keeps improving at the rate it has, I don’t see why either stays protected forever.

They’re the moat today, but I’m not confident they will be in five years.

Copies are cheap

Over the last week we saw the entire Adobe suite get cloned, and it’s actually pretty good…. if you build something new and exciting and someone can clone it in no time. A weekend, maybe less… Depends if they have the available AI resource. The interface, the features, the copy on the landing page, all of it, every jot and tiddle.

But existence isn’t the same thing as use, a clone nobody uses is worth nothing. The thing that’s hard to copy was never the code, rather everything around it.

So the levers that matter now are distribution, exclusive data or access, network effects and established operations. Who already has the customers. Who has the data nobody else can get. Who has the relationships, the trust, the boring operational machinery that keeps a business running.

That flips a lot of startup wisdom on its head. For years the story was that a small team with a better product could beat the incumbent. Now the small team can build the better product in a month, and the incumbent can copy it in a week and push it to a million existing users. Anything novel without real backing will either get lost in the noise or copied and shipped through networks with far better reach. Meanwhile slopware copies of everything will flood the market.

Established platforms become more valuable, not less, because they own what AI can’t replicate.

One good example of this in action is one of my very own projects…

I am a space nerd, I built my own 3d solar system with the intent on making it the most photorealistic experience on the web and I created 3dsolarsystem.online in a space that was already saturated with 3d solar system projects, with the most famous being NASAs very own “Eyes on the Solar System”.

On my own I’ve been able to create the most immersive impressive visual experience available, that surpasses even NASAs deep pockets. But I’m literally one guy…. my dedication to experience and quality and my knowledge of the industry has allowed me to rocket to 5th place in google for “3d solar system

As you can see, I get a ton of returning users and over 10k users visiting per month, the insane thing is the quality I am producing was entirely created by AI with my curation, I did not write a single line of code for my solar system…

“We’ve heard this before”

The obvious response to all of this is that we’ve been here before… Fair.

Compilers were supposed to shrink programming, so were high level languages, then frameworks, then cloud, then open source. Every time, engineers got more productive, software got cheaper, and demand grew so fast that headcount went up anyway. Economists call it the Jevons paradox. Make something cheaper and people use more of it.

Maybe that happens again, cheaper software means more software gets built, and some of that will need people.

But every previous wave made engineers more productive while still needing an engineer at the keyboard. A compiler doesn’t decide what to build, a framework doesn’t write the feature, this wave is different because it’s starting to need fewer people in the loop at all.

The reckoning

I have no idea on the timeline, but I think a reckoning is coming for the software industry, I wrote about this last years in regards to the white collar collapse of the economy…

Engineering teams will need fewer people as systems become more capable and autonomous. Hiring, already a trickle in plenty of places, slows further. Leavers don’t get replaced. Teams shrink gradually, not in one dramatic round of layoffs but quietly, one unreplaced seat at a time, until you’ve got a skeleton crew keeping things running with AI doing the heavy lifting.

Large generalist agencies will largely go the way of the dodo. Their whole model is selling engineering hours, and engineering hours are exactly what’s getting cheap. The ones that survive will be specialists, or the ones that become something other than agencies.

Agencies that make money by billing juniors at a markup and staffing projects with ten people, when two people with AI can do the work are just not going to exist anymore… Brochure sites, basic apps, simple integrations. Clients will either do these themselves with AI tools or pay very little for them. Hourly billing as the main pricing. If the work takes a tenth of the time, billing by the hour cuts your own revenue. Agencies stuck on it will shrink or die.

Customers will still buy accountability and transfer risk though, a non-technical founder or a company without an engineering team doesn’t want to orchestrate agents, they want someone to blame if it goes wrong. That’s a big part of what agencies have always sold.

That having been said, large organisations will keep using approved suppliers for years regardless of what’s technically possible.

Agencies get AI too however, the smart ones will shrink headcount, switch to fixed price or outcome-based pricing, and keep the margin.

What keeps me up at night

The software industry will adjust, what worries me more is the wider picture.

As productivity rises, companies need fewer workers. Fewer workers chasing fewer jobs means wages don’t need to stay high, unless you’re doing a genuinely sought after specialist job. The gains won’t vanish, they’ll get concentrated. More output, fewer people sharing in it, more of the profit going to whoever owns the platforms and the capital.

That having been said… cheap software doesn’t make housing cheaper, It doesn’t make energy cheaper, It doesn’t make food cheaper, those are the things we’re all paying more for than ever before, and none of them get easier because an app can be built in a weekend.

So you can end up in a world where the stuff that’s easy to make is nearly free, and the stuff you actually need to live keeps getting more expensive, while your income is under more pressure than it’s been in a generation.

What the hell can you do about it?

I don’t have a grand answer, I’m as worried about the future as the next person who’s informed, so I’ll just say what I’m doing.

I’m keeping the day job and protecting that income for as long as I can and upgrading it if I can. Not out of loyalty to a job title, but because steady income is what buys you time to try everything else…

I run my own consultancy on the side. Partly for the money, partly to keep my skills sharp and up to date but mostly so I’m not dependent on a single employer’s view of what I’m worth.

I build my own products, cheaply and quickly, because AI has made testing an idea cost almost nothing. If something doesn’t land, I’ve lost very little and actually gained, because I enjoy creating things... If it does land, it might become one of those things that’s hard to copy like the community I’m building around my solar system.

I’m building a financial cushion, so if things shift faster than I expect I’ve got room to move instead of being forced into whatever’s available.

None of that is about predicting the future correctly. It’s about not being dependent on one job title when the ground moves beneath your feet.

We’re surfing the wave right now, and it’s a good ride cowabunga dude. But how long until it overtakes us, and what comes after, I genuinely don’t know, do you?

BTW if you’re a space nerd checkout my solar system https://3dsolarsystem.online and consider purchasing my screensaver ;)