If you’ve spent time on X lately, I’m sure you’ve seen the refrain…

SaaS is DEaD!!1!11!

It looks like people are swapping their subscriptions for vibe-coded replacements, tools built exactly for them, with no monthly bill attached.

(There’s even a website for it now, CanIVibeCodeIt, where you type in the app you’re paying for, and it tells you whether you could plausibly build it yourself, and how much you’d save doing it.)

But is any of this working at scale?

Hard to say.

Some of it is real, some of it is noise. People venting anxiety about subscriptions and jobs on a timeline because it feels like doing something about it.

(I did the exact same thing a few months ago, so no blame at all.)

Back in February, I wrote about whether it was the end of SaaS as we know it, after a couple of BlackTwist customers canceled and pointed at Claude and their own custom dashboards as the reason.

Yes, I spiraled a little.

My conclusion at the time was that customization would be the deciding factor, that no packaged product could out-personalize what someone builds for themselves.

I don’t think that was wrong, but I think I was asking the wrong question.

Whether you can build something that fits you perfectly is no longer in doubt. With Claude Code or Codex, you probably can, for almost anything.

What matters is whether it’s worth your time.

And that comes down to leverage: what do you get back for the hours you put in?

Build something once and it serves a thousand people, that’s leverage. Build something once and only you will ever use it, that’s just work, and whether it’s worth doing depends entirely on what you’re replacing.

Take my own expense tracker as an example.

It’s a Google Form that dumps into a spreadsheet, with a pivot table on top. Ugly, zero automation, and I’ve wanted to replace it with a real app for years already.

Building it myself would take, let’s say, two working days. Call it 16 hours. At a conservative $30 an hour, that’s $480 of my own time, plus another $100 to $200 in Claude Code or Codex to actually ship it.

Make it $600, and that’s assuming no scope creep, which never happens on side projects.

Now compare that to what I’d actually be replacing.

My current setup costs $0 and needs zero maintenance. No subscription to cancel, no bug to fix at midnight, nothing to migrate when a plugin or library breaks.

Building my own version wouldn’t save me a dollar I’m currently spending. It would just cost me $600 and possibly a weekend to replace a workflow that already works.

If I were paying for an off-the-shelf app instead, say $19 a month, the math shifts but not enough to flip the answer.

At $19 a month, it takes close to three years of payments to equal what building it myself would cost upfront. Three years of not touching a database schema, not debugging an edge case at midnight, not thinking about it at all.

(Usually a good trade.)

Of course, none of this means building your own tools is a mistake.

If you want to spend a weekend making something because it’s fun, or because you want to get better at using Claude Code, that’s reason enough on its own.

But if the goal is efficiency, the math rarely works out in favor of building.

A tool you build yourself doesn’t just cost you the hours to ship it but also every hour after that too, every time it breaks, every time your workflow changes and the tool doesn’t.

So…

These days, before you build anything, ask yourself: “What exactly am I replacing, and what does it actually cost me right now?”

If the answer is “nothing, it already works,” that usually settles it.

The same logic probably applies to SaaS at scale, too. Not every subscription deserves to survive AI making DIY easier. Some of them were overpriced or overbuilt for what people actually need, and those will lose to a Notion template and an agent. But not every subscription is replaceable just because it’s technically possible to replace it now.

Sometimes, the tools worth keeping are the ones where someone else is eating the maintenance, or the edge cases, or the 2 am bug.

Sure, we can build almost anything we want at this point.

But remember…

Just because you can, it doesn’t mean you should.