Song-saem: From Planner to Developer with AIBlog home →
What Can You Build with AI?5 min read한국어로 읽기

How Have Website Databases Changed? From Hosted MySQL to Supabase and AI

Website databases have gone from MySQL on a hosting account to cloud servers, managed databases and backend services like Supabase, and today AI tools can connect to a database and help with the work. Here is what each step made easier, what you still decide yourself (what to store, access rules, backups, cost), and how to narrow permissions when you connect AI.

How have website databases changed? Many sites used to keep a PHP website and a MySQL database together on a hosting account. Then came the cloud, which made servers easy to rent, managed databases that take care of installation, patches and backups, and backend services like Supabase that bundle login and file storage too, so the choices widened. The older ways didn’t disappear; they are all still in use. As things got easier, deciding what to store, who sees it and how to restore it is still up to the builder.

How was data stored before?

PHP with MySQL was a long-popular pairing. PHP built the pages, and MySQL stored member details, posts and orders. You rented space from a hosting company, uploaded the website files, and looked at the database in an admin screen such as phpMyAdmin. FTP here is a tool for moving website files, and it has a different job from a database management tool.

That doesn’t mean site owners did all the server work themselves. Hosting companies often managed the server and the database software. How far they covered website software updates, data management and backups, though, depended on the contract. I covered the security side of that era in Why Did Old PHP Websites Get Hacked So Often?.

What did the cloud and managed databases change?

ApproachExampleWhat the service handlesWhat you look after
Hosting accountPHP + MySQL, phpMyAdminServer and database software, depending on the contractWebsite software updates, data, checking backup coverage
Cloud serverAmazon EC2 (launched 2006)Hardware and facilitiesUpdating the operating system and any database you install
Managed databaseAmazon RDS (launched 2009)Operations such as installation, patches and backupsData structure, how it’s used, permissions
Backend serviceFirebase, SupabaseDatabase, login, API, file storageTable design, access rules, checking backups

The cloud made it possible to rent the servers you need quickly over the internet. But renting a server easily and managing what runs inside it are different jobs. If you rent an ordinary virtual server, updating the operating system and the database you installed is still on you. That is why managed databases, which take on more of the database operations, became important, letting builders focus more on how their data is structured and used.

Databases moving from servers you ran yourself to managed services
Databases moving from servers you ran yourself to managed services

What does Supabase bundle together?

Backend services like Firebase and Supabase widened the scope further. Besides the database, they provide login, an API (the channel for sending and receiving data), file storage and more.

That was the case in Build Your Own Scheduler too. I made a table for schedule data, added login, and set a rule so that each signed-in person sees only their own schedule. It takes away much of the burden of building all of that from scratch.

Supabase is built on a database called PostgreSQL. Rather than two competing databases, think of Supabase as a service that bundles PostgreSQL and other features so they’re easy to use. And today, hosted MySQL, self-run servers, managed databases and backend services are all still used side by side.

How are serverless and AI different?

A “serverless database,” as services like Neon use the term, doesn’t mean there is no server. It means less server management for you, with the service handling more of the resources. Features like pausing automatically or scaling up, and the pricing terms, differ by service.

AI is a separate story. AI isn’t a new kind of database. It has become a tool that connects to an existing database to look at its table structure, query data and draft commands to change it. Supabase’s MCP connection plays this role, letting AI coding tools connect to Supabase and help with the work. It’s one more way to handle a database.

What do you still decide yourself?

What to decideWhat to check
What to store and access rulesThe service won’t decide for you what information to keep or who can see it
BackupsSupabase recommends that free-plan users regularly export their data and keep it separately. On paid plans, know how long automatic backups are kept and what they cover. A database backup doesn’t necessarily include the actual photos and attachments in file storage
Code vs. dataPushing your code to GitHub doesn’t bring back the schedules and customer records you entered
Free limits and costThe free tier may be enough at first, but terms can change as data and usage grow

What matters is not picking the cheapest service, but deciding how your records are kept and how you’ll get them back when needed.

Even when a service runs it, you keep the key to your data
Even when a service runs it, you keep the key to your data

What about connecting AI to a database?

Asking AI only to explain is different from giving it the right to change a real database. When you first look at the structure, connect it to a development project and start read-only if you can.

Also tell apart saying “just read it” from actually setting the tool to read-only. Narrow the project it connects to and the tools it’s allowed in the settings, and when a change is needed, check the command to be run first. You could start with a request like this:

“Look at the tables and access rules in the connected development project. Don’t change any data yet; explain the current structure and anything that needs checking in plain words. When you suggest a change, first show me the SQL you’d run and how it would affect existing data.”

Adding a new feature can be checked with test data first and then applied to the live service. Have your backup and restore method ready before that.

Showing AI only as much data as it needs
Showing AI only as much data as it needs

In short

  • Website databases went from hosted MySQL to cloud servers, managed databases and backend services like Supabase, and these approaches are all still used together.
  • A serverless database doesn’t mean there is no server; the service takes on more of running the resources.
  • AI isn’t a new database; it’s one more tool for looking at and working with an existing one.
  • What to store, who sees it, and how to back up and restore it are still the builder’s decisions.
  • When you connect AI, start with a development project and a read-only setting, and check the commands before they run.

At ONE-PEX, building the business system FROX, we think the more company records a service holds, the more it matters to decide from the start where the database lives and who looks after backups and permissions.

Sources (checked 2026-10-27)

Related reading: Why Did Old PHP Websites Get Hacked So Often? · Part 2: The tools you need to build a web app

  • #데이터베이스
  • #DB변천사
  • #MySQL
  • #Supabase
  • #클라우드DB
  • #웹앱만들기
Song-saem · Sanghwa Song, CEO of ONE-PEX

A software planner who became a developer after two years of working with AI five to twelve hours a day. Today I build FROX, an AI-powered business system, at ONE-PEX.

More posts on the blog home →