Software Is Still Eating the World. There's a Storm, and the Opportunity Is Here.

There's never been a better time to be in software. I know that sounds backwards in a year of layoff headlines and "AI is killing the junior developer" hot takes. Stay with me.

The job that's disappearing isn't the developer job. It's a particular version of the developer job. The version where a smart person gets handed a spec and asked to translate it into code without questioning whether the spec is right, whether the customer actually wants this, or whether there's a better way to solve the underlying problem. The industry has had a name for that role for thirty years. Code monkey. We invented the slur ourselves, for our own people, because we built an industry that needed humans to act like computers and didn't have computers good enough to do the work.

We do now. Good.

The part of the job that forced humans to behave like machines is getting handed off to actual machines. The part that requires judgment, taste, customer empathy, and the call on what to build when nobody's specced it yet stays with us. That's where the value compounds. If you're a developer who already works that way (and most of the good ones do), AI is rocket fuel. If you've been pinned down doing rote spec-execution because that's what your team or your manager asked of you, the next few years will feel like a promotion.

Now, about those layoffs.

I'm empathetic to anyone living through one. They're brutal, and they're real. But let's be honest about what's actually driving them, because the dominant story doesn't hold up. Big tech didn't lay off hundreds of thousands of engineers because AI suddenly made them redundant.¹ Big tech laid off hundreds of thousands of engineers because from 2020 through 2022, the industry went on a hiring binge fueled by free money: zero-interest-rate policy, pandemic cash, and the conviction that growth would compound forever.² Meta and Alphabet roughly doubled their headcount over that period.³ Microsoft added eighty thousand jobs. Amazon's footprint doubled. Then interest rates moved, the ad market wobbled, and the math stopped working. The correction was coming with or without ChatGPT.

"AI is replacing them" is just a cleaner story than "we over-hired by tens of thousands." One makes shareholders nervous about the past; the other makes them excited about the future. Guess which one ends up in the press release.

The Easy Half

In August 2011, Marc Andreessen wrote an essay in the Wall Street Journal called "Why Software Is Eating the World."⁴ The thesis was that software companies were going to take over chunks of every industry, and the people running those industries should be terrified.

He was right. Retail got eaten by Amazon. Taxis by Uber. Hotels by Airbnb. Payments by Stripe and Square. Media by everyone. The categories Andreessen pointed at are now owned, top to bottom, by software companies that didn't exist or barely existed in 2011.

But he was only right about the easy half.

Software ate the parts of the economy with clean workflows and digital-native substrates: the businesses where replacing a phone call with a webform was enough. What software hasn't truly eaten, fifteen years later, is the half of the economy that runs on judgment and human relationships. Healthcare, construction, agriculture, defense, government, logistics, education. Some of these are digitized in the literal sense (there's a screen in every hospital and a CRM in every dealership), but digitized isn't the same as eaten. In a lot of these sectors, software has actually made the human bottleneck worse. Your doctor doesn't have a clipboard problem. She has a thirty-page electronic medical record problem, with five specialists' notes, a half-dozen lab results, and a medication history sprawled across three systems that don't talk to each other. The data exists. No human can ingest it in the seven minutes she has between patients. Software brought a flood; it didn't bring a riverbank.

The reason these sectors held out for so long is that they sit behind a wall I wrote about last month in The Scaling Wall.⁵ The wall has three layers. The first is structural: services where the labor is the product can't scale by adding tools. The second is economic: for most of these sectors, custom software was only marginally better than Excel at orders of magnitude more cost, so organizations rationally stayed on the spreadsheet. The third is structural again: a fifty-thousand-engineer technology giant cannot possibly understand forty-seven industrial verticals from the inside, and it cannot hire enough engineers to brute-force them either. Communication complexity, the Ringelmann effect, and the patience required to ship something that works in a niche all conspire against scale.

So the wall stayed up because nobody small enough to climb it had the tools to build what was on the other side.

Now they do.

AI doesn't make the wall faster to climb. It lowers it. The non-judgment work that surrounds every act of skilled human labor (the documentation, the data wrangling, the integration friction, the rote pattern-matching) is exactly what AI absorbs best. As that overhead falls away, the height of the wall between what a small focused team can deliver and what a sector actually needs collapses with it. A four-person team with deep knowledge of surgical scheduling, municipal permitting, or agricultural insurance can now build the software that sector has been begging for, with leverage that five years ago would have required a five-hundred-person company.

That's the world we're walking into. Not less software, more. Not fewer builders, different ones. The walls didn't fall on their own. They got shorter because the overhead dropped, and the small teams who can climb them just got their leverage.

Where The Work Went

Here is the part of the story nobody is telling.

While big tech was correcting, two parallel things happened. One of them made every headline. The other made very little noise. The headline thing was the layoffs. The quiet thing was that the cost of starting and scaling a software business cratered, and a generation of small teams walked through the open door.

Look at the U.S. Census Bureau's Business Formation Statistics.⁶ New business applications surged more than 24% to over 4.3 million in 2020, then climbed to over 5 million in 2021 and have stayed there every year since. The pre-pandemic baseline was 2.5 to 3 million applications a year. We are now running at roughly twice that rate, and we've held it for five consecutive years. The dominant narrative says the economy is shedding tech workers; the actual data says the country is forming more new businesses per quarter than at any point in modern history. Both are true. We just only talk about one of them.

Look at the app stores. Look at GitHub. Look at the indie hacker revenue datasets.⁷ The number of independent operators shipping working software, generating revenue, and serving real customers has gone vertical. A solo founder with a Claude subscription and a Stripe account can do what a fifteen-person team did five years ago. There aren't fewer software companies. There are dramatically more of them, and most of them are tiny.

Why does this matter for the article we're writing? Because the layoff narrative is built on a hidden assumption that the only legitimate place for a software career is inside one of five large tech companies. Strip that assumption out and the picture changes completely. The seats inside those five companies got fewer; the seats outside got vastly more. Some are at small startups. Some are at indie studios. Some are inside non-tech companies that are finally building real software for the first time. Some are at one-person operations whose founder used to be an engineer at a big tech company until they figured out they could build their own thing. The total number of seats in the building went up.

I have been arguing for fifteen years that small teams win.⁸ I believed it when capital was cheap and headcount was easy and the conventional wisdom said you needed a hundred engineers to ship a serious product. I believe it more now. The economic conditions just stopped making it a contrarian view and started making it the obvious one. A team of three with AI agents and deep customer empathy beats a team of three hundred with everything else. That's not a hot take anymore. That's where the work is going.

So here is the picture the headlines aren't drawing. Big tech got smaller. The rest of the software economy got dramatically bigger. The work didn't disappear; it got distributed. The opportunity didn't shrink; it spread out. And the people who lean in now, especially the small teams and the early-career builders willing to try, are walking into the most interesting moment software has had in decades.

That's the macro picture. In the next post, I'll get specific about what the new operating model actually looks like, with a metaphor I think captures it cleanly: the bridge of the Enterprise.