The SaaSpocalypse

Is the SaaSpocalypse real? Will AI coding tools really destroy the moats of today's technology firms, plunging the economy into recession? Or is AI a hoax, another attempt to string along investment in a long-running tech cycle? The truth is that the SaaSpocalpyse is coming, but not in the sudden, calamitous way that AI commentators envision. Instead, just as industrial firms rearranged their processes around the assembly line, some technology firms will wholeheartedly rearrange their product development around AI. The first to do so will reap enormous, structural benefits, and the stragglers will fade into irrelevancy.

Let's get a lay of the land around our two camps, the AI doubters and AI doomers. The doubters focus on the moment-to-moment hysterics of the AI boom: the daily drama of Sam Altman and Dario Amodei, or whatever Jensen Huang is up to. For example, Gary Marcus rightfully calls out the AWS outages caused by AI, or difficulties with long-term maintenance of a real-life codebase. Thankfully for us, there are less naive ways to use AI besides letting it take over your legacy code; it isn't a condemnation of AI writ large, just in the way it's being used. The doubters seem to be right on the small details, and yet have consistently failed to foresee the larger trends, at this point for over three years.

On the other hand, the AI doomers correctly identify AI disruption on the horizon. A great example is the substack post from Citrini Research, which seemingly triggered a market sell-off after publication in late February. The problem with the doomer outlook is the same as the doubters, in that they both rely on broad, imprecise readings of what AI is. Let's work off a snippet from the Citrini post to see what I mean:

The DoorDash moat was literally "you're hungry, you're lazy, this is the app on your home screen." An agent doesn't have a home screen. It checks DoorDash, Uber Eats, the restaurant's own site, and twenty new vibe-coded alternatives so it can pick the lowest fee and fastest delivery every time.

It's probably correct that the Uber Eats mobile app could be vibe coded, but an app is a tiny sliver of the entire business. It requires heaps of data to be useful: a list of restaurants, their menus and prices, nearby drivers and estimates of delivery time, and so on. This data must be deliverable to a user's device, which means it must be stored somewhere, which means there must be a mechanism to transmit it in the first place. And this is just for the user trying to order food; the other two types of users, deliveristas and restaurant owners, require an equal or even richer product offering. Finally, we must mention that products are not static, but evolve over time. Every software engineer knows that to update an existing product, without breaking anything, requires the care and attention befitting a heart surgery. Because AI commentators from either camp aren't involved in the day-to-day operations of building and maintaining a software-based product, these details are invisible to them.

Clearly, a tech company's operations cannot be delegated away to vibe coders. We can already see from the above example that there is a difference between writing code and maintaining a product. I believe that the key to assembly-line software lies in pushing this concept to the extreme, in melting down existing SDLC processes and recasting them into product-to-code pipelines, purpose-built to let AI run at full throttle, and let humans take on only the most complex, creatively demanding tasks. The result is a reorientation, away from software engineering, and towards software design. This website is dedicated to the hard work of making this a reality.

In the future, there will be new firms who run their software operation with this mindset, and it will offer them massive advantages against incumbents. Primarily, these new firms can operate with a lower headcount, translating to more competitive pricing, but there are also knock-on effects. When there are less cooks in the kitchen, everything goes faster: less meetings, less bureaucracy and corporate politics, less chance for departmental fiefdoms to take hold. Teams invest more time into their features, while delivering them faster. Markets constantly shift and form new niches, these firms will win the race to fill them because they can refactor their products faster than the competition.

So the SaaSpocalypse isn't the sunset of the white collar firm so much as it is the dawn of a new one altogether. The better analogy is a SaaS extinction event -- the mammals will go on to greater things, and the dinosaurs will become fossils.