Notion's linked databases still win

Notion is having one of those revealing media moments where two stories land at once and accidentally explain the whole product. One writer finally “got” Notion after building a system with linked databases. Another ditched a paid Notion setup after Claude Code learned their note habits and handled the routine work in plain markdown. Both stories are true. But they’re true about different layers of the product. ### What’s the actual split here? The split is between note-taking and system-building. If your job is mostly dumping thoughts somewhere, AI agents are getting very good at that layer. Claude Code can work inside a folder of markdown files, search across them, rewrite them, and reorganize them with natural-language prompts. That makes a lot of personal knowledge work feel less tied to any one app. ### Why did the anti-Notion story resonate? (howtogeek.com) Because a lot of people use Notion like a prettier text editor with sidebars. The XDA piece is basically about that trap. The writer had a messy notes section, then found that an AI agent working on local files could infer patterns and reduce the manual sorting. If the pain you feel is “I spend too much time cleaning up notes,” an agent that cleans up notes is a direct substitute. (anthropic.com) ### So what do linked databases change? They turn pages into records. That sounds small, but it’s the whole game. In Notion, a database entry is also a page, which means each item can hold rich content while still behaving like structured data. Then you can filter, sort, group, and show the same underlying database in different views. A linked database lets that same source appear elsewhere without duplicating it. ### Why are relations and rollups the hard moat? (msn.com) Because they let one object know about another object. A task can belong to a project. A project can roll up task status. A customer can connect to purchases. A content idea can connect to a publishing calendar. That is closer to a lightweight app builder than a notebook. Once your workspace starts behaving like a little relational system, replacing it with “AI plus folders” gets much harder. (howtogeek.com) ### Couldn’t AI rebuild that too? Sometimes, yes — but the catch is reliability. An agent can infer patterns from files and automate repeated cleanup, but inference is not the same thing as schema. A schema is explicit. It tells the system what counts as a project, what counts as a task, and how they connect every time. That matters when you want dashboards, consistent properties, and repeatable workflows instead of a smart assistant improvising from context. That’s an inference from how Notion’s database model and Claude Code’s file-based workflows differ. (notion.com) ### Where does AI actually hurt Notion? At the top of the funnel and in the casual middle. If someone never graduates from simple notes, AI makes Notion easier to leave. The old lock-in was “my stuff lives here.” The new reality is “an agent can work on my stuff anywhere.” That weakens retention for users who were never really using Notion’s deeper model in the first place. ### Where does Notion still win? It wins when the workspace is the product. (notion.com) Founders, operators, and heavy personal-systems people are not just storing text — they’re building interconnected objects with views, dependencies, and summaries. Linked databases are still one of the cleanest ways to do that without writing software. AI can help operate the system, but the system still needs shape. ### Bottom line? AI is attacking Notion’s easiest use case first — generic note-taking. (howtogeek.com) But linked databases still look like the durable core. If your workflow needs relations, rollups, and reusable structure, Notion is not just a place to write. It’s the database.