Attention Is All You Need
1 comments
Claude can help shorten the dev process, but as you suggested without a good architecture and intentionality, things will eventually break if you vibe-code a large solution. And since you didn't build it, you won't understand how to fix it.
SaaS might have had a lull, but it should hit a resurgence with so many enterprising people being able to code full modern websites with advanced functionality for very low cost, without a small team of devs.
And yes, attention is indeed mandatory for success.
SaaS might have had a lull, but it should hit a resurgence with so many enterprising people being able to code full modern websites with advanced functionality for very low cost, without a small team of devs.
And yes, attention is indeed mandatory for success.
On the surface, it sounds reasonable. But after a decade of building software professionally, I can tell you software is never just written once. It it re-written over and over again.
It always feels like you could just spin something up and be done. If that were true, software companies would ship a product once and call it a day. Instead, they have engineers working on the same product for years. That’s because of two things:
1. Use cases evolve, feature sets grow. 2. Every product has edge cases and bugs. Even if Claude gets incredibly good, the moment you scale or expand scope, new problems show up.
And sometimes nothing is “broken,” but your understanding of how people use the product keeps evolving. The way something is used today won’t be how it’s used tomorrow. As developers, we’re constantly updating that mental model.
Case in point: CRM is probably the most shat-on category in SaaS. Everyone thinks it’s trivial. “It’s just a database with a UI, I could build this over a weekend.”
And sure, you can build a version of it. Contacts, deals, maybe a pipeline view.
But the moment a real team starts using it, everything breaks in subtle ways.
- Sales wants custom workflows. - Marketing wants attribution. - Leadership wants reporting that somehow reconciles everything.
Now you’re dealing with permissions, integrations, data drift, sync issues, performance under load, and a hundred edge cases you didn’t think about.
It stops being “just a database” very quickly.
And once you’ve built software for a while, you realize you’re always one feature away, one bug away from “done.” So even with Claude, building software is still a full-time job.
If you try to replace every SaaS tool you use with something internal, you won’t save time. You’ll just spend it chasing the next tweak, the next edge case, the next missing feature.
I believe even after throwing all the tokens in the world at a problem, to build a remarkable product that works well, ATTENTION IS ALL YOU NEED.