Medium's new AI assistant lives inside the draft, not a separate tab
Medium launched TK on November 9: type @TK in the editor and an AI assistant answers, edits or pulls up something you wrote weeks ago. It also makes writers disclose what AI touched.
Medium announced a new app on Monday, November 9, called TK. Small detail, big trivia. “TK” is the placeholder editors use for “to come,” a fact to fill in later, and it’s spelled with a K instead of a C specifically because that letter pair almost never shows up elsewhere in English, so a proofreader’s eye catches it fast. Medium borrowed the convention and built an app around filling in the gaps it marks.
What TK actually does
The core idea, per TechCrunch’s report and Medium’s own launch post, is that the AI assistant lives inside the document instead of a separate chat tab. Type “@TK” followed by a request and the app answers inline, whether that’s a quick fact (how many days in a given month) or pulling back a paragraph you saved to the app while you were still working out an idea. TK ships with specific skills too: editorial feedback, headline feedback, and matching a draft against a Medium publication’s style guide.
It can also reach outside the app. TK connects to a user’s own AI agent through an MCP server or a Claude skill, so an assistant that already holds context about your projects and positions can plug into what you’re writing, rather than starting from zero each time.
Nothing forces you to use any of it.
What TK is, and isn’t, at launch
Here’s the practical picture, since TechCrunch and Medium’s own post describe it slightly differently in places.
| Feature | At launch |
|---|---|
| Platform | Web only at launch per TechCrunch; The Verge reported it’s also live on Android and iOS for existing members |
| File format | Imports and exports Markdown, but isn’t a native Markdown editor |
| Encryption | Not encrypted, since the app is built for sharing drafts and feedback |
| Price | Free for anyone with a Medium account |
| Agent connection | MCP server or Claude skill, optional |
The disclosure angle
Medium is leaning hard into a side of the AI-and-writing debate that’s been contentious all year. Literary prizes have faced scrutiny after suspected AI-assisted entries, and outlets including The Atlantic have covered writers quietly leaning on AI tools without disclosure. Medium’s launch post doesn’t dodge that. “AI is a technology that has the power to amplify human intelligence and it uses the written word as a foundation,” the company said, framing AI as something that makes writing more important, not less.
The disclosure piece is the part I think matters more than the assistant itself. TK lets authors attach provenance notes that spell out how AI was involved in a piece, down to the prompting and source material, according to Medium’s post. That’s a real, structural answer to the “was this written by AI” question that plagues sites like this one, instead of a vague AI-use policy buried in an About page. Whether writers actually use the disclosure feature honestly is a different question nobody can answer yet.
My read
The AI-inside-the-document approach is the more interesting bet than the AI-assistant part, which every writing app has shipped some version of by now. What Medium is actually testing is whether “this is how much AI touched this piece” can become a normal thing a writer shows a reader, the way a photo credit or an edited-by line already is. If that catches on anywhere, a publishing platform with Medium’s scale is a more plausible place for it to start than a browser extension nobody’s heard of.
Note on sources: I read Medium’s own launch post for TK in full and TechCrunch’s report on the same launch. Where the two differed on a detail, like whether mobile is supported at launch, I flagged it in the table above rather than picking one.
Sources