I Stopped Building Webflow Sites for a Month. Here Are the 3 Apps I Built Instead.

Why I Stepped Away From Webflow

I did not touch a Webflow site for a month, and it was not because I gave up on it. Webflow is still an excellent tool for fast, well-designed marketing sites. I spent the month building apps instead, and three of them are now live: Suboptima, LUMIO, and VYXO.

The reason was simpler: I wanted to know whether the way I work with AI could stretch beyond websites. I had already used Claude Code to build and audit sites. The next question was whether I could use the same approach to build products where a user signs in, data is saved, and the software has to keep working after I hand it over.

Suboptima: A Subscription Tracker

Suboptima helps you track, manage, and optimize all your subscriptions in one place. It shows what you pay each month and year, when each subscription renews, and where you might save money.

You add subscriptions manually, with no bank or email connection. The app supports more than 150 currencies and keeps totals separate instead of mixing them. It sends email reminders the week before a renewal, lets you filter by category, exports to CSV, and includes an AI advisor that suggests savings, such as cutting overlapping services or switching to a yearly plan.

It is a Next.js web app on Supabase, and it taught me my first real lessons about authentication and security hardening.

LUMIO: One Reading Lesson a Day

LUMIO is a free app for web and Android with the tagline “Less scrolling. More understanding.” Each day it gives you one focused 15-minute reading lesson, chosen for you from classic public-domain books. Every lesson ends with a few takeaways, and each idea is traced back to its source.

You pick your subjects when you start, from areas like history, psychology, and science. Your reading place is saved, and there are no quizzes or guilt-driven notifications, just a gentle streak.

It is built with React Native and Supabase, with background workers that prepare lessons. The database schema lives in versioned migrations, and I validate every SQL file with PostgreSQL’s real parser before applying it. On a Webflow site, a CMS mistake is visible immediately. With a database, a mistake can sit quietly until it hurts.

VYXO: A Private AI Workspace

VYXO is a free, open-source AI workspace that runs on your own computer, with the tagline “The AI that stays in the room.” It uses Ollama to run models locally, so chats and files stay on your machine. There is no account, telemetry, or subscription, and chat works offline once a model is installed. It currently supports Apple Silicon Macs.

It is aimed at people who cannot casually paste confidential material into a cloud tool, such as lawyers, developers working on unreleased code, consultants, and researchers. Projects keep their own goals and instructions, proposed edits appear as diffs and are written only after you approve them, and memory only changes when you say so.

The trade-off is honest: local models are smaller than the strongest cloud models. For some work, privacy matters more than raw power, and building VYXO made me think hard about that choice.

What Actually Changed in How I Work

I think in data before layout. Webflow trained me to see a page as structure and style. Apps start earlier: what is stored, who can read it, and what happens when it is wrong.

I think about trust. Whether it is a renewal email, a saved reading position, or a file that never leaves the machine, every app made me decide what the user is trusting me with.

I write tests and document decisions. I rarely needed them for site work. In the app projects, typechecks, linting, tests, and short notes about why something works the way it does became part of the routine.

I use the AI agent differently. For sites, I mostly asked it to build and fix. For apps, I use it more like a pair programmer: it proposes, I check the reasoning, and I keep project notes up to date so it does not drift. An AI agent does not remove the need to understand your system. It raises the cost of not understanding it.

What I Would Tell Other Webflow Developers

You do not need to leave Webflow to try this, and you probably should not abandon it for work it does well. But if you have been curious about apps, your background is more useful than you think. You already understand structure, responsive behavior, content modelling, and how to listen to a client.

What you need to add is comfort with data, authentication, and failure. The best way to learn a backend is to build one that has to work.

Webflow taught me to think in structure. Apps taught me to think in data. Together, they are the difference between building something that looks right and building something that keeps working once real people depend on it.