Medium launches TK, betting blogging belongs inside groups
Medium launched TK for web, Android and iOS. The free app adds AI-assisted notes, private topic libraries and a bid to rebuild publishing beyond Google search referrals.

- TK is free on the web, Android and iOS; a native desktop app remains forthcoming.
- CEO Tony Stubblebine called TK Medium's most important launch since 2017.
- Early TK Libraries range from a few members to a few hundred, with no public pages.
- Medium moved most content behind a subscription paywall in 2017.
- TK does not track pageviews or reward authors for attracting traffic.
By Elena Vasquez
Medium has launched TK for free on the web, Android and iOS, adding a note-taking app built around private publishing groups rather than dependence on Google search referrals.
TK supports daily journal entries, longer-form writing and AI-assisted organization. Medium also promises a native desktop app soon. No release date accompanies that promise.
Medium presents the app as a way to encourage regular writing and sharing with friends and fans. The broader thesis puts group chats at the center of blogging's future.
What changed since 2017?
During a pre-launch interview, Medium CEO Tony Stubblebine said: “I don’t think we’ve actually had a launch of anything important since 2017.” He tied its significance to the year then-CEO Ev Williams stopped chasing advertising and put most Medium content behind a subscription paywall.
Williams wrote at the time that the advertising-driven internet was broken and did not serve people. Stubblebine said the company had left the attention economy. TK extends that earlier subscription move into a model based on closed networks and collaboration.
The comparison sets a high bar for a product without disclosed adoption, retention or revenue targets.
Takeaway: TK is a strategic pivot, not yet a proven business shift.
How does TK use AI?
TK offers a sparse, Evernote-like interface for daily notes and longer pieces. Writers can direct an AI model to search existing notes, organize material or respond to a draft with feedback.
Typing TK summons that assistant. The shortcut draws on a common placeholder used by writers, because the letters T and K rarely appear together, making the pairing easy to find later.
Stubblebine invokes Unix pipes while describing how large language models could handle repetitive work in creating and managing notes. The launch details do not name the models, explain data handling or set limits on AI assistance.
That silence matters for a product built around personal writing and group collaboration. Users receive no stated basis for evaluating model accuracy, privacy or control.
Takeaway: The AI feature is available; its model and governance remain unspecified.
How do private Libraries work?
Sharing works much like sharing a Google Doc. A user can send links or email a draft, and collaborators can write, comment and add emoji reactions.
TK also provides topic-based collections called Libraries. Stubblebine calls the concept private blogging. Early examples contain book projects, recipe exchanges and poetry drafts.
Groups range from a few people to a few hundred, according to the early examples reviewed for the launch, and their activity has no public pages. TK does not track pageviews or reward authors for attracting traffic.
“This is a community that’s going to give you feedback and encouragement,” Stubblebine said. “It’s not about the drive to be noticed.”
He describes Medium as performative because it asks writers to expose themselves publicly. In a post-Google publishing environment, he considers that proposition nearly impossible.
The announcement does not specify access controls, moderation rules or membership limits for Libraries. It also does not say when private-community features will reach the broader Medium platform.
For publishers built around subscriptions or member communities, the relevant comparison becomes retention and paid conversion, not pageview growth. TK offers no such metrics yet.
Takeaway: Libraries replace public reach with feedback. Their membership rules and growth remain undefined.
Can TK reduce Google referral risk?
Medium still relies on Google for a large share of its traffic. Stubblebine argues that AI companies take content regardless of legality, while search engines and chatbots send less traffic back to publishers. He calls that condition Google Zero and treats it as an existential risk.
He says he warned Google that reducing referrals would hurt Medium and Google, because Medium's roadmap would move publishing into private networks. TK needs no Google referrals, but Medium's existing properties still depend on them.
Here, the confirmed product design is straightforward: TK offers private publishing without search referrals. The claimed market shift rests on Stubblebine's assessment, without a named traffic series or independent study in the interview.
The announcement describes no Google search or ranking change. It discloses no partnership or referral agreement involving Google.
Takeaway: TK reduces referral dependence by design. It does not remove Medium's existing exposure to Google.
How could private publishing make money?
Stubblebine points to newsletters as a way to charge readers for private or paywalled material. He also proposes packaging selected content from a Library, such as Recipe Box, for paid access.
He goes further, asking whether AI agents could pay for access to the strongest human writing. That proposal remains hypothetical. TK has no announced agent API, pricing structure or licensing terms.
Stubblebine argues that the generic web is losing intelligence and that future AI needs a layer of human intelligence. TK might supply that layer outside the Google algorithm, in his view.
The proposal has two potential customers: people seeking curated material and machines seeking licensed content, but neither route has a disclosed price, rights framework or conversion forecast.
Takeaway: The revenue case rests on newsletters and untested ideas for paid collections and AI-agent access.
What should readers monitor next?
Monitor four items: the native desktop app, integration of private-community features into Medium, adoption of TK Libraries, and any disclosure of AI models, data policies or monetization terms.
These checkpoints will separate a durable publishing network from an attractive interface. Medium says private features will reach the broader platform soon, but gives no timetable. TK's free access and Google-free referral model also leave open questions about subscriber conversion and the cost of AI processing.
The next test is not whether TK can generate posts. It is whether private groups retain writers without public traffic once the model moves beyond the app.
Takeaway: The launch proves availability, not adoption, retention or revenue.
via theverge.com (Source)
More from Elena Vasquez
Show full bio
Correspondent covering people, launches and day-to-day practice in Search rankings and algorithm updates, generative engine optimization at SERP Journal.
221 articles