A 7-Year Typora Run Ends: Closed Plugins, Image Upload Failures, and 600 Articles Head to Obsidian
Typora's refusal to expose a plugin API or marketplace leaves it unprepared for AI-assisted writing, while Obsidian's open plugin ecosystem is already absorbing AI features. For writers with large markdown archives, the migration cost is real: switching editors means deciding whether the back catalog becomes an AI-trainable knowledge base or stays locked in a monolithic desktop app.
Typora's closed, monolithic design blocks AI-era extension. Official releases ship without a JS plugin API or a marketplace; the most popular community plugin framework works by hacking Electron entry files, injecting JavaScript through DOM hijacking, so any major DOM refactor breaks it. CSS themes are the only official extension point.
Image uploads to a GitHub bed through PicGo fail frequently due to network flakiness. Failed uploads downgrade to local images with no automatic retry, forcing manual re-pasting for republishing; long articles with many images freeze a 32G Mac. The file tree is hidden by default and only appears on hover.
The writing view itself is the deeper problem. Full-width lines violate typography research: Chinese reads best at 30-45 characters per line, roughly the central 40-50% of a widescreen display. Obsidian and MarkText both put the text in a narrow center column by default, while Typora's focus and typewriter modes don't solve the peripheral distraction. After testing both alternatives, Obsidian's plugin store, AI integration, instant global search, and file tree offer a path to turn the 600-article archive into an AI knowledge base; MarkText fixes the writing mechanics but its stalled maintenance, high memory, and input bugs rule it out for the AI workflow. The decision is not about which editor is lighter; it's about shifting from a writing tool to a knowledge system.
Typora's closed architecture is not the problem by itself; the absence of an official plugin API matters because it prevents the community from extending AI features in a stable way. The DOM-hacking plugin approach is a sign that demand outstrips official extension points.
The typography argument reframes the writing experience: wide lines cause more cognitive fatigue than any missing AI feature, so editor design decisions about line width directly affect long-form writing productivity.
Obsidian's Vault model changes the economics of markdown files: 600 articles stop being a static archive and become a searchable, AI-queryable asset, which alters the migration decision far more than minor UX differences.
MarkText shows how an open-source clone can fix UX gaps, but its maintenance history undermines trust for a long-term writing tool, even when the feature set is superior.
After seven years and more than six hundred posts, this farewell is indeed a bit hard to part with. For pure Markdown writing, Typora's real-time rendering is still smooth. Curious which tool the author moved to later.