The Week I Kept Saying the Same Thing Five Different Ways
From the “Entrepreneurship without the fantasy” desk
Every developer I talked to this week — and most of the content I made — kept circling the same uncomfortable idea: the thing you built your instincts around might be the thing most exposed right now.
That’s the thread. Let me run it out.
Your Rails instincts are a liability
If you’ve been building SaaS with Rails for a few years, you’re fast and clean. You know the patterns cold. Active Record, resource-based architecture, REST conventions — you can ship in your sleep.
Here’s the problem: those instincts are optimized for exactly the category of software AI kills first. Rails is extraordinarily good at putting a structured interface in front of data so a human can make a decision. That’s the superpower. It’s also the vulnerability.
AI agents don’t need the interface. They don’t click buttons or read dashboards. They call the API and do the thing. The mental model that made you productive maps almost perfectly onto the SaaS category most exposed right now.
That’s not a reason to abandon Rails. It’s a reason to question what you’re building with it.
Workflow tools vs outcome systems
Most SaaS products are workflow tools. Someone shows up, looks at data, makes a decision, clicks a button. That’s the product. It worked for twenty years because humans were doing the work.
There’s a different kind of product: an outcome system. It doesn’t guide you through a process. It skips the process and delivers the result.
Concrete example: a dashboard that shows a seller their inventory trends is a workflow tool. A system that analyzes the data, decides what to reorder, and places the order — that’s an outcome system. Same problem. Completely different product. One needs a human to show up. The other doesn’t.
This is the question worth asking about whatever you’re building: am I helping someone do a task, or am I removing the need to do it at all? Those aren’t the same thing anymore.
The honest test
One question I’d ask every Rails developer building SaaS right now: if an AI agent had direct access to your database and your APIs, would it still need your product?
Sit with that honestly.
For a lot of products, the answer is no. The product exists to help a human navigate data — read a report, review a status, click approve. If there’s no human in the loop, there’s no need for the product.
This isn’t hypothetical. Customers aren’t cancelling because they found a better tool. They’re cancelling because an agent just does the job. No dashboard. No workflow. No monthly subscription.
If the answer to that question makes you uncomfortable — good. That discomfort is useful information.
The ceiling is gone and you’re still in the basement
You have access to the same AI as a billion-dollar company. And a lot of developers are using it to build a better to-do list.
I’m not being glib. Every developer I talk to knows something is wrong. Nobody quite knows what to do next. So they slap AI on whatever they were already building and call it a pivot.
But the parts exist right now to build things that were science fiction five years ago. Real diagnostics. Home health monitoring. Early disease detection. Not conceptually — as a product roadmap, today, with one person and a laptop.
AI didn’t just make coding faster. It changed what one person can actually attempt. The question isn’t “what’s the next AI opportunity.” The question is — now that the ceiling is gone, why are you still building in the basement?
Your moat was friction all along
Developers talk about moats. Network effects. Integrations. Switching costs.
Here’s the uncomfortable truth: most of what we called a moat was just friction. And AI dissolves friction.
Think about what switching away from your product actually requires. Export the data. Learn something new. Migrate the workflow. That’s not loyalty. That’s inconvenience. When an AI agent can read your schema, map your logic, and rebuild the core of what you do in an afternoon — inconvenience stops being a moat.
I’ve watched products with serious ARR assume they were protected. Not because they were defensible. Because switching was annoying. That’s a different thing.
The actual moat is narrow: proprietary data, unique logic, actions that require trust you’ve already earned. That’s the list. Most products aren’t on it.
Mine included, if I’m being honest.
The through line
Five pieces of content this week. One idea underneath all of them: the comfortable assumptions that made you a productive builder are now the first things worth questioning.
Not because everything is broken. Because the game changed while you were shipping.
The developers who come out ahead aren’t the ones who move fastest. They’re the ones who stop and ask the right question first.
Phil Smy has been building startups for 20+ years, including a Shark Tank appearance. He lives in rural Japan and writes about entrepreneurship without the fantasy.