I Replaced a $120/Year SaaS in 20 Minutes with an LLM — And You Probably Can Too
The Pragmatic Engineer’s experiment shows why “good enough” SaaS products are about to face extinction.
The Core Insight
Gergely Orosz had a $120/year subscription to Shoutout.io, a simple service that displayed testimonials on his website. It worked fine for four years. Then the billing section broke, customer support sent a broken link, and Gergely asked himself a dangerous question:
“Could I rebuild my use case with an LLM?”
Twenty minutes later, using Codex, he had a fully functional replacement. No third-party dependency. No annual fee. Same visual result.
The SaaS is dead. Long live the prompt.
Why This Matters
We’ve heard the theory before: AI will kill SaaS by making it trivial to build custom software. Most of us dismissed it as hype. After all, products like Workday handle compliance, Salesforce manages complex workflows, and these aren’t things you can “just prompt away.”
But here’s what Gergely’s experiment reveals: the theory is correct for a specific, massive category of SaaS — products that don’t provide ongoing value.
His testimonial widget had one job: display some quotes. Once set up, it sat there. No compliance updates. No dynamic data. No continuous value-add. The only thing keeping him subscribed was inertia.
That inertia just got a lot easier to overcome.
The calculus has changed:
– Before LLMs: Rebuilding = hiring a developer or spending weeks learning to code
– After LLMs: Rebuilding = 20 minutes with Codex and enough technical judgment to verify the output
For developers, this is liberating. For SaaS companies selling static features, this is existential.
Key Takeaways
“Set it and forget it” SaaS is vulnerable. If your product doesn’t continuously earn its subscription, you’re one frustrated customer away from being replaced by a prompt.
Broken windows accelerate churn. Gergely’s migration started because billing was broken. In the AI era, customers have a new option beyond “complain and wait” — they can simply rebuild.
Developers have a massive advantage. Non-technical users can probably replicate this, but it takes longer. Devs can port SaaS features in a lunch break.
The buy vs. build equation has shifted. “It’s cheaper to buy” used to be the default answer. Now it’s “it depends on how static the feature is.”
Micro-SaaS acquisitions may become less profitable. The model of buying small SaaS tools and extracting subscription revenue without investment relies on customers staying. That assumption is weakening.
What This Means for SaaS Builders
If you’re building SaaS, ask yourself:
Does my product provide ongoing value? Compliance updates, real-time data, dynamic insights — these are defensible. Static features are not.
Am I earning my subscription every month? If a customer could replace your core feature in an afternoon with AI, they eventually will.
What happens when my billing/support breaks? In the past, customers complained and waited. Now they have an exit ramp.
Is my moat actually a moat? “It would take too long to rebuild” is no longer true for many use cases.
Looking Ahead
This doesn’t mean all SaaS is dead. Complex, compliance-heavy, continuously-evolving software has a strong future. Workday isn’t getting replaced by a prompt anytime soon.
But the long tail of simple, static SaaS tools? The testimonial widgets, the link-in-bio pages, the basic form builders? Those are now competing against the 20-minute LLM rebuild.
The winners will be products that deliver continuous, irreplaceable value. The losers will be the ones surviving on inertia — because AI just eliminated the friction that kept customers from leaving.
For developers, this is actually exciting. Every frustrating subscription you’re paying for a simple feature is now a candidate for the “can I rebuild this with AI?” experiment.
You might be surprised how often the answer is yes.
Based on analysis of “I replaced a $120/year micro-SaaS in 20 minutes with LLM-generated code” by Gergely Orosz (The Pragmatic Engineer)