❧ If you’re not someone who has coded much, I’m not sure you can appreciate how merited the more reasonable hype around AI is. Things that used to take me weeks to do now take me hours. And more importantly, it’s allowed me to customize my computing experience more than I ever imagined would be possible.
No matter how much you love a piece of software, there is always a thing or two you wish it would do differently. More typically the software available to us is nothing like what we would want for a task, especially in terms of a UI that makes sense to us.
Now that coding agents have gotten quite good, however, I don’t hesitate to just write from scratch the app I want to use, with the exact UI and functionality I want, even if I already have a good enough commercial alternative. Todo lists, editors, whatever. And I will also make bespoke tools that do something for which I’m the only customer.
It’s remarkable what I can get done, and it makes me think that this power is going to have to show up in the productivity stats eventually. Even if it’s just software that AI supercharges.
Yet I can also see the downsides. Since I didn’t write this code myself, I don’t really understand it and don’t know how good it is. That may come back to bite me. Given my coding proficiency, however, I can’t believe it’s doing worse than I would do. That said, the real alternative to a vibecoded app is not one I would write by hand, but instead simply using the hardened commercial app. I’m also taking on a maintenance and upkeep tax whose size I can’t now appreciate.
And then there’s the addiction piece. This can’t be an original insight, but using AI lately has reminded me of Addiction by Design, the book on machine gambling about which Stably and I recently podcasted. Sitting alone at a machine, dropping in ‘credits’ and ‘tokens’ of inscrutable value, pulling a lever, getting back variable rewards, repeating this turn after turn, and ultimately losing track of time and oneself. In a way all flow is like this, but the jury may still be out on whether in the long run vibecoding induces the ‘good’ kind of flow that produces mastery and a sense of agency, or the ‘bad’ kind that surrenders agency to the machine’s rhythm.
I tell clients that code, like nearly any tool we adopt, is both an asset and a liability. That is to say that it's something that enables us to do something new or something more quickly or something more easily, but it's also something we have to maintain and could break and tends to do so when least convenient. The open source world has done a lot to bring down the liability side of the ledger by essentially making less and less code custom. You may build a custom website, but you build it on WordPress with themes and plugins, essentially mixing modular components together to get what you need. You may build your own CMS, but you don't author your own database solution, scripting language, or web server. You may decide to built a new kind of web server, but you're building on Linux, not your own OS or kernel. Essentially at whatever level of the stack you're focused on, you can build on top of the work of others and that work is updated on a regular cadence in a way that generally doesn't break what you've built on top of it. I think AI coding is going to see this pattern generally hold, but we're going to see so many more contributions, modules, and plugins at every layer, but especially at the frothy top of the stack where code turns into user interfaces and connections with the services we use daily. WordPress themes are going to be replaced by systems made for agents to customize while staying modular and keep code coherent and easy editable. WordPress itself may lose some market share to bespoke CMSs that you spin up in day. But at some level the customization and the vibe coding stops. How many people are going to vibe code a database or a reverse-proxy cache? Not just because these projects are complex, but because there will still be things that are simply not worth your time to build and not worth the maintenance headache even if the investment has fallen from 1000 hours to 100 or even from 10 to 1. Off-the-shelf works right now, not after an afternoon chatting with Fable or Astra. I also think liability has to be thought of in terms of human liability. No one ever got fired for buying IBM for a reason—it wasn't considered a negligent decision. Similarly, no one will get fired for continuing to pay for SaaS solution that's too cheap to fuss about, but they may get fired for attempting to code their own CRM only to have it lose a single sale that cost their firm 10x the SaaS subscription cost for the entire year. Buy a service from someone else makes the service's fuck up their problem, especially if it's a vendor that your peers trust and has reputation for being the solid play. Even though you chose them, the problem is their fault. In other words, we pay for functionality, but we also pay to pass off liability on to someone else, to pass of maintenance on to someone else, and to not have to think about things. I think that will hold for a lot of things, but the difference the AI makes at the margins where the customization happens will still be profound. In just better integrating the SaaS solutions I pay for, AI is worth every dime I spend on it. Benioff will get his ARR for the foreseeable future and, if anything, his systems will become more valuable as a stable platform from which AI agent can operate. You don't have to be too situationally aware to understand this. I do believe there is a something to the addiction angle you bring up. AI is super fun to use. It's thrilling. I watch the machine do thing that used to take me weeks or months. Now done in an afternoon or in days of back-and-forth prompting and re-prompting. I think there's not only a danger in the addiction, but a danger in not thinking systematically about what you're doing. Because one-off things can be done so quickly, you may be tempted to treat problems as one-off and unique when they are actually solvable through a repeatable process. Taking a step back and remembering that software in general and AI in particular can automate, rather than merely accelerate, a task is important to avoid wasting time mucking about with the chatbot. In short, I think gains from AI are going to happen, but it won't be because "SaaS is dead" or because everyone will be building a full stack of apps. Instead, it's because the sort of digital last mile is finally going to be finished. It's the turrets and finials that will be built by end users with AI while the load-bearing walls and foundations will still be produced by open source projects and for-profit software ventures that look something like they do today at least in the way the way to relate to the broader ecosystem. Yes, they may be using AI as well, but they will still exist as independent nodes in the broader world of software. There will not be a software forge in every backyard simply because economies of scale, specialization, shifting liability, and other forces will still hold.
2026.09.22