Vibe coding's real cost: upkeep nobody budgets for

The spreadsheet of anyone deciding to build with vibe coding usually has two columns. On one side, the price of a SaaS subscription or a studio. On the other, a few dollars in AI platform credits. The math looks obvious. But a line is missing.

In February 2026, Jason Lemkin, founder of SaaStr, published an account of what people are actually building with vibe coding. Lemkin knows what he's talking about: SaaStr has more than 12 AI-built apps in production. And he shares what it costs to keep them running.

What SaaStr built with AI

Over a year, SaaStr built its apps mostly on Replit. A few numbers from the piece:

• AI VC Suite neared 1 million uses, with more than 3,000 pitch decks reviewed.
• AI Mentor had more than 100,000 conversations with founders.
• The AI-driven review of speaker submissions evaluates more than 4,000 applications and, according to Lemkin, saves more than US$180,000 a year.
• Prototypes are ready in 20 to 60 minutes. An internal org chart tool shipped in three days.

It's a real success story. And that's exactly why the maintenance side deserves attention.

The missing line in the spreadsheet

Lemkin and Amelia, SaaStr's head of AI, each spend 30 to 60 minutes a day maintaining the apps. Every day.

He's blunt: “This isn't set-it-and-forget-it software.” And he adds that going beyond the prototype “isn't a weekend project, it's a commitment.”

Do the math. Two people, up to an hour a day each. That's up to two hours a day from two of the most expensive people in the company, the founder and the head of AI. Over a month, that's dozens of hours. None of them shows up in the comparison between “a US$200 monthly subscription” and “a few dollars in credits.”

What maintenance includes

Maintaining an app built with AI, part of the cost of keeping a system live, isn't just fixing bugs. It includes:

1. Platform changes. APIs change, libraries update, integrations break.
2. Data that drifts. An app in use piles up cases nobody foresaw.
3. Security. Lemkin cites a vibe-coded app that leaked 72,000 images. Security isn't solved once; it's watched.
4. User requests. A product people use generates change requests every week.
5. Performance. What worked with 10 users can stall with a thousand.

Lemkin also admits the limits of what he knows: on maintaining all this without professional developers over the long run, he says “honestly, I don't know yet.”

Why SaaStr can pull it off, and why that doesn't apply to everyone

The context matters. SaaStr has a founder who has lived and breathed SaaS for years, a person dedicated to AI, and apps that mostly serve its own community. If one of them goes down for an afternoon, there's a cost, but it's manageable.

Now picture a solo founder with a product customers pay to use every day. That same daily hour of maintenance comes out of time for selling, supporting customers and fundraising. And when something breaks on a Friday night, there's no second person to share the problem.

The SaaStr case proves you can keep AI-built apps in production. It also proves that maintaining them costs experienced people's time, every day.

Where vibe coding makes sense

The piece isn't against vibe coding, and neither is this article. Lemkin shows where it shines: quick prototypes, internal tools, event sites, replacing simple SaaS. And he makes clear that complex, critical, deeply integrated software isn't being replaced yet.

The question for the founder isn't “vibe coding or not.” It's: who will spend the daily hour on maintenance, and what is that hour worth?

What this means for you

Before you decide to build your product yourself with AI, redo the spreadsheet with these lines:

1. Build cost (credits, tools, your time).
2. Hours a day of maintenance, multiplied by what your hour is worth.
3. The cost of a security incident involving customer data.
4. Opportunity cost: what you stop doing, like selling, while you maintain the app.

If the math still works after that, great. If it doesn't, you found out before spending.

How we do it at the Studio

Post-launch care is part of what we deliver. Every system ships with an AI agent that runs daily, flags issues, suggests improvements and checks that the data makes sense. It's the daily hour of attention SaaStr does by hand, running automatically and alerting whoever needs to act. On the human side, our in-house dev team tracks those alerts in the Ongoing evolution service and responds within 2 hours during business hours.

If you already have a product live, built with AI or not, and want to stop spending an hour a day keeping it running, that's what our ongoing evolution and system maintenance service is for.