Logseq's fundamental unit is the block — a bullet point in an outline. Blocks can link to each other, creating a graph of connected text. This works beautifully for journaling and note-taking, but it doesn't scale to structured data. A task is a block with a checkbox. A contact is a block with a name. There's no way to say "show me all tasks due this week" without manually maintaining tags and queries.

Synap's fundamental unit is the entity — a typed data object with a profile, properties, and relationships. A task has a status (todo/in-progress/done), priority, and due date as first-class fields. A contact has an email and company. Because the data is structured, views can render it meaningfully: tasks on a kanban board grouped by status, contacts in a table sorted by company, events on a calendar.

In Logseq, every connection is manual. You type [[Page Name]] to create a link. You add #tags by hand. You write queries in a custom syntax to create live views. The result can be powerful, but it requires continuous effort. If you stop maintaining the system for a week, new information sits unlinked and untagged.

In Synap, AI handles the linking. Capture a research paper, and AI detects it's an article, extracts the author, links it to related entities by topic, and files it with the right tags. Forward a client email, and AI creates or updates a contact entity and links it to relevant projects. The knowledge graph grows automatically, even when you're not thinking about structure.

Logseq stores everything as markdown files on your local disk. This is genuine ownership — you can open the files in any editor. The trade-off is that markdown files can't represent structured data well. Properties live in frontmatter, relationships live in wiki-links, and there's no referential integrity or query optimization.

Synap stores data in a dedicated PostgreSQL database. You get the same ownership — self-host it, export with pg_dump, connect any SQL client — but with the power of a real database: typed columns, foreign key relationships, full-text search, vector embeddings for semantic search, and server-side AI processing. It's your data, in a format that scales.