The Future Of Software Is Small
The dominance of SaaS platforms in business and pleasure today is a cyclic one. If you went back in time to the 1980's and told them that 21% of the industry was using the same CRM platform they'd probably nod serenely, knowing that nobody ever got fired for buying IBM.
You might surprise them, though that these platforms were all available over the internet and run in a web browser (admittedly, they wouldn't know what that is because it hadn't been invented yet). Jump forward to the late 90s and the idea that over 60% of Americans got their news, weather, chat, and online forums from a single provider would also have probably been met with an impressed nod towards the survivability of America Online.
Of course, the dominant platforms of today aren't the same ones that seemed eternal thirty-odd years ago. Things do change, and I think the cyclic nature of these changes is worth ruminating on. I tend to subscribe to the idea that the driver behind these changes is mostly an economic, rather than technical, one. Wal-Mart didn't decide to in-house their IT and build a world-class logistics platform because it was a good idea, they did it because it was cheaper to do so than the alternative -- at least, in net. AOL didn't fall out of favor just because the World Wide Web opened the doors to a world of publishing and self-expression, but because the economics of building the experience you wanted on the Internet outweighed the network effect of email and chat. These aren't 100% pure rational effects, obviously. I'm not an economist, there's a lot more here than the surface level, but I think it's worth remembering that these trends were shaped, ultimately, by tradeoffs around time.
Time, Tides, and Abstractions
It's been quipped that 'all code is technical debt', which is true. Every LOC you add increases your maintenance burden, and expands the surface area required to operate software at scale. Small issues or flaws in interface design, architecture, abstraction boundaries, etc. will accrete over time like barnacles on a ship's hull. As an industry, we've made this worse through decades of convenient abstractions (especially around hardware!), trading away understanding for ever-faster delivery of features. Little wonder, perhaps, that people keep blowing off their foot with surprise charges. We've made it very easy for anyone to build economically useful stuff, but we've traded away our ability to really own what we build, and how we run it. For what it's worth, I think it's good that we've made hardware easy! It's good that we've made it easier to program. There's value in abstraction, there's value in making the internet and applications more accessible. What I think is a mistake, though, is that we've become over-reliant on platforms rather than on the underlying protocols. We're building, as an industry, at the wrong layer of abstraction.
This loss of ownership shows up most dramatically in discovery. The easiest example is, of course, the Apple App Store -- in most markets, if you're not on the App Store, your application simply doesn't exist. While the glory days of Facebook Games may have passed, a significant amount of many products strategy goes straight thru Meta's policy and APIs. If you're making games, better hope payment processors don't get pressured into thinking your content is smut or else you're shit out of luck because -- again -- you don't exist without Steam or Itch or any other marketplace. Similarly, if you're a B2B application, enjoy the thrilling experience of the AWS/GCP/Azure Marketplace and pray to god that a bored PM doesn't decide that they're gonna directly compete with your solution. Want to innovate in the world of CRM? Better hope whatever you're doing integrates with Salesforce!
To simplify, we've made it easier than ever to do something but we've made it harder to really interpret what's going on. We've built all of these abstractions, but they're all built on top of middlemen who would like their 30% cut, please. It's not great!
The Part About AI
Remember where I said that the economic incentive behind these earlier shifts was driven by time tradeoffs? If you were Wal-Mart, it was worth your time to build your own retail logistics platform so you could put the screws to suppliers, optimize your logistics, and eventually come to dominate American retail. If you were an internet user, why pay for AOL when you could get Netscape or IE for free? and go to all sorts of pages? Did AOL have the Hamster Dance? I think not. I would argue that platforms rise and fall based on this implicit (or explicit) time economics. When time is expensive, platforms do better; When time is cheap, they do worse.
AI makes time very, very, very cheap. It's not unreasonable to expect that within ~5 years, we'll have consumer-grade hardware with onboard capabilities that rival current state-of-the-art models (Claude 4, etc.) that are also faster and cheaper than those models are today. This is an absolute sea change in terms of capability at an OS level; Custom applications will become the norm, not the exception. Why try to grapple with fitting my life into Notion (or whatever) when I can just have the computer build me bespoke applications that work on all my devices and are catered to my precise needs? Why do I need planet-scale infrastructure to share baby photos with, like, 5 people?
This goes for business as well; Why do I need a legion of Salesforce consultants to make their shit work with my shit when I can just have the AI write all the reports I'll ever need? Same thing for HRIS, ERP, and dozens of other fields.
The great vibe shift in software is going to take place on these battlegrounds - not the locked-down platforms of today, but the ocean of data management and access of tomorrow. In this, I believe we'll see small start to win again. Small, custom programs for individuals, families, teams -- with data sharing, discovery, and management built on open protocols. Things like ATProto and some of the other interesting outgrowths of crypto are lights in the darkness here, imo. There's other stuff too -- Tailscale, for instance, and the ease of creating small private networks. There's more to be done; Discovery is a huge one, indexing is another, private content is a third. We also, critically, need standards work and protocol work to be elevated in both speed and visibility. This is an area where we, as technologists, can have a huge impact -- it's time for us to act like it and act accordingly.