Artificial intelligence is now capable of writing.
Enter a topic in a short sentence, and AI can generate a title.
It can summarize research materials, organize an outline, and produce thousands of words in seconds.
It can revise sentences or imitate the writing style of a particular author.
Writing has become easier than ever.
But that progress raises a new question.
If anyone can use AI to produce a finished article, what value remains in writing something ourselves?
Online publishing platform Medium has introduced a new response to that question.
On November 9, 2026, Medium unveiled TK, an AI agent-powered tool designed for writing and thinking.
In an official blog post introducing the product, Medium CEO Tony Stubblebine presented a philosophy that differs from the prevailing direction of the AI industry.
While much of the industry is investing in a future where machines do more thinking on behalf of humans, Medium wants to invest in a future where people think more and write more.
Stubblebine described the company's approach as a bet on human intelligence.
That does not mean Medium is rejecting AI.
On the contrary, TK incorporates a range of AI capabilities.
Writers can call on AI while drafting, retrieve previously written notes and manuscripts, request editorial feedback, and evaluate potential titles.
They can also connect external AI agents to their personal knowledge repositories.
The difference lies in the purpose of those capabilities.
Rather than focusing primarily on having AI generate complete articles, Medium wants to help people develop their own ideas more effectively.
This approach has important implications as generative AI becomes increasingly widespread.
As the cost of producing written content with AI falls, the volume of available writing could increase dramatically.
In that environment, the competitive advantage of a publishing platform may shift away from how much content it can produce toward how effectively it helps people develop valuable ideas and distinctive forms of expression.
TK is a product built around that possibility.
1. Why Did Medium Create a New Writing App?
Medium was founded in 2012 by Twitter co-founder Evan Williams.
It developed into an online publishing platform where individual users, writers, and professionals could publish their work and connect with readers.
One of Medium's defining characteristics has been the relative simplicity of its publishing process.
Writers can make their work available to the public without building their own websites or managing complicated publishing systems.
However, the place where writing is published is not necessarily the place where ideas are developed.
Writers often use multiple tools to gather information and record notes.
They organize ideas, prepare drafts, and revise their work several times before publication.
Publishing a completed article on Medium may be the final stage of a much longer process.
With TK, Medium wants to move into the stages that come before publication.
Rather than remaining simply a destination for finished writing, the company intends to participate in the process through which writing takes shape.
Stubblebine explained the relationship through a simple metaphor.
TK is the gym where writers develop their abilities.
Medium is the stage where they present the results.
That metaphor captures the company's business strategy.
In TK, writers record ideas and develop their thinking.
On Medium, they publish finished work, reach readers, and potentially earn compensation.
The company is extending its existing publishing platform into the creative process that precedes publication.
2. The Name TK Comes from a Longstanding Writing Convention
The name TK comes from a convention used by professional writers and editors.
In English-language journalism and publishing, TK is a placeholder indicating that information will be added later.
It stands for "to come."
For example, a journalist who cannot remember an exact date while drafting an article may leave TK in the relevant sentence.
The information can then be verified and inserted during editing.
Medium has transformed this longstanding convention into an AI interface.
In the TK app, writers can invoke AI by typing @TK while working on a sentence.
Consider a writer who needs to check how many days September has.
Previously, the writer might have interrupted the drafting process and opened a search engine.
With TK, the question can be asked directly from the writing interface.
The writer can also request a particular paragraph from something written the previous day.
The significance extends beyond convenience.
The design seeks to reduce the interruptions caused by switching between tasks.
Moving between writing software, search engines, note-taking applications, and AI chatbots can disrupt the flow of thought.
TK aims to make AI available when needed without requiring writers to leave the page.
Instead of pulling humans into a separate AI workspace, Medium is bringing AI into the space where humans already think and write.
This suggests that where AI functionality is placed within a product may become an important source of differentiation in user experience.
3. There Is a Difference Between AI Writing for You and AI Helping You Write
A wide variety of generative AI writing tools already exists.
Users can provide a topic and ask AI to produce a complete article.
They can also request revisions tailored to a particular style, length, or audience.
These capabilities can significantly reduce the time required to produce written content.
But they often place the AI-generated result at the center of the experience.
TK takes a different approach.
It treats writing itself as a process of thinking.
In its official announcement, Medium explained that writing is difficult because thinking is difficult.
The company emphasized that eliminating the difficulty of thinking is not its objective.
Instead, it wants to preserve the experience of developing ideas through writing while making relevant information and feedback easier to obtain.
Within this philosophy, AI plays a different role.
It is not intended to replace the writer.
It functions as an assistant that helps identify overlooked problems, examine alternative structures, and improve the quality of the writer's work.
Of course, users may still employ TK's AI features to generate substantial amounts of text.
The product's philosophy cannot determine how every individual will actually use it.
Nevertheless, Medium's stated direction is clear.
It wants human thinking and writing to remain central, with AI serving a supporting role.
4. Why Does Medium Emphasize the Accumulation of Ideas Rather Than Writing Alone?
TK's second major function involves using previously written material to support new thinking.
People record ideas at different moments in their lives.
A question may arise while reading a book.
An idea may be recorded as a voice note during a commute.
An unfinished manuscript may be saved for later.
The difficulty is that these materials are often scattered across multiple applications.
Even when someone has recorded a valuable thought, they may not remember where it was stored.
Earlier ideas may also remain disconnected from new writing projects.
TK is designed to address this problem.
Users can import articles they previously wrote on Medium.
They can also bring in material from note-taking tools such as Bear and Obsidian.
Mobile applications and widgets provide ways to record text and voice notes.
Through these features, Medium wants users to accumulate their writing and notes within a personal knowledge base.
The important objective is not simply storing information.
It is making previously recorded ideas useful again.
A note written several months earlier might become relevant to an article being drafted today.
A topic explored repeatedly across multiple essays might eventually develop into a serialized project or the outline of a book.
TK uses AI to help users discover those connections.
The system is designed so that each new writing project can draw upon the ideas and records accumulated through earlier work.
5. Could Personal Writing Become a Knowledge Asset in the AI Era?
Medium introduced an especially important concept in its announcement.
The company suggested that personal intelligence, rather than general intelligence, may be a scarce resource.
Generative AI models have already been trained on enormous amounts of general knowledge.
They can produce sophisticated responses to questions about history, economics, technology, and science.
So what distinctive asset can an individual bring to AI?
Medium's answer is the person's own recorded thinking.
Which problems have mattered to them?
What arguments and perspectives have they developed?
Which projects have they pursued?
Which sentences have they revised, and which ideas have they abandoned?
Such records can contain a person's intellectual history—information that a general-purpose AI model may not otherwise know.
TK seeks to make that history a foundation for new AI-assisted work.
As users accumulate writing and notes, AI can retrieve earlier ideas and identify relationships among them.
In this structure, personal records become more than an archive.
They become resources supporting new thinking and creative activity.
However, maintaining a large collection of notes does not automatically lead to better reasoning.
Inaccurate information and outdated judgments may also be retrieved and reused.
When AI incorporates earlier records, users must therefore consider the source, date, and accuracy of the information.
Nevertheless, the idea that an individual's intellectual history can become a valuable data asset deserves attention.
6. How Can Scattered Notes Become a Personal Encyclopedia?
TK also offers functionality for connecting scattered records into more structured knowledge.
According to Medium, search results, topics, tags, and libraries can be summarized into wiki-style pages.
These pages may contain overviews of major themes, lists of related ideas, and links to original documents.
Consider someone who has spent several years writing about AI and education.
Initially, the material might consist of unrelated essays and notes recorded at different times.
With TK, the writer could organize those materials into a more coherent knowledge structure.
How is AI changing education?
What role should teachers play?
How should educators understand students' dependence on AI?
Connecting earlier records around such questions could provide the foundation for new articles or research projects.
This functionality relates to the broader field of personal knowledge management, or PKM.
Established note-taking applications use document links, tags, and folders to organize information.
TK aims to automate parts of that organizational process using AI.
This helps explain Medium's emphasis on avoiding the need for users to learn complicated organizational systems before they can begin.
The idea is that people should be able to record thoughts freely while AI helps retrieve, connect, and structure them.
However, automatically generated summaries and classifications are not always accurate.
AI could misunderstand the context of an older note or combine different arguments as though they represented a single position.
Those limitations will require careful management.
7. Why Does Medium Emphasize Human Feedback Alongside AI Feedback?
TK provides two different forms of feedback.
The first comes from AI.
AI can evaluate article titles and examine writing structure.
It may suggest revisions based on the style guides of particular Medium publications.
It can also help identify opportunities to develop several pieces of writing into a book.
The second form of feedback comes from people.
Medium has emphasized that social feedback remains especially important.
To support that experience, TK includes a feature called Social Libraries.
Users can share their writing with others and receive feedback.
At launch, three public libraries were available.
Public Journal is intended to help users develop a regular writing habit.
Get Feedback on Your Writing allows writers to share drafts and receive comments before publication.
Write a Book supports people working toward turning manuscripts into books.
These features extend writing beyond a purely individual activity.
AI can provide immediate responses.
But understanding how actual readers interpret a piece of writing or what emotions they experience is a different matter.
Persuasiveness, empathy, and the relationship between a writer and an audience can develop through human interaction.
Medium is attempting to preserve that dimension of writing while incorporating the convenience of AI-generated feedback.
8. Why Does Medium Want to Reveal the Writing Process in an Age of AI-Generated Content?
One of the most notable aspects of TK is its approach to AI transparency.
Medium said TK would support disclosure of information about which parts of a text were written by humans and which were produced by AI.
The company also described a broader direction involving visibility into a work's sources and creation process.
Its stated ambition includes making prompts and reference materials available for inspection.
This differs from conventional AI-generated content labels.
Much of the existing discussion about AI disclosure has focused on whether a piece of content was generated by AI.
For example, a label might indicate that AI produced part or all of an article.
But actual AI-assisted writing is more complicated.
AI may have been used to search for information.
It may have suggested a title.
It may have rewritten a few sentences.
A writer may have produced the entire original draft and then used AI for editing.
Alternatively, AI may have written most of the text, with a human making relatively limited revisions.
Reducing all these processes to a single label indicating "AI use" can obscure important differences.
Medium's approach aims to make aspects of the production process visible, rather than disclosing only whether AI was used in the final result.
This could give readers a more detailed understanding of how a piece of writing was created.
However, an important distinction remains.
Disclosing the writing process does not automatically guarantee the factual accuracy or quality of the finished work.
An article can contain false claims even when its AI usage is fully disclosed.
There are also questions about which parts of the writing process should be made public and how the completeness of those records can be verified.
TK's transparency features may therefore help support trust, but they should not be treated as technology that independently guarantees a complete and accurate record of authorship.
9. Will Demonstrating Human Contribution Become More Important Than Simply Labeling AI Content?
Understanding human contribution is becoming an important issue in AI-assisted writing.
The mere fact that AI was used does not make all AI-assisted writing equivalent.
In one article, a human may have identified the central problem, gathered the evidence, and used AI primarily as an editing tool.
In another, AI may have handled most of the work, from selecting the topic to producing the initial draft.
Both articles may carry the same AI disclosure label.
Yet their creation processes are substantially different.
Those differences could influence perceptions of originality, credibility, and the author's contribution.
Medium is attempting to introduce a different approach by emphasizing transparency throughout the writing process.
Where did the central argument originate?
Which notes and references informed the work?
What prompts did the writer use?
Which sections involved AI assistance?
If such information is available, readers may gain additional grounds for evaluating the article beyond its final wording.
However, greater disclosure does not necessarily produce higher perceived value.
Readers may prioritize different characteristics.
Some may value accuracy and usefulness above all else.
Others may place greater importance on direct human expression and personal experience.
The effects of AI transparency on reader trust, willingness to pay, and content evaluation therefore remain questions requiring empirical research.
Nevertheless, it is significant that Medium is beginning to treat the creation process itself as a product feature.
AI disclosure may evolve from a notice attached after publication into a function integrated throughout writing and publishing.
10. Why Is Medium Building an AI App After Criticizing AI Training Practices?
Medium's AI strategy may initially appear contradictory.
The company has previously criticized the collection and use of online content by AI companies for model training.
It referred to that position again in its TK announcement.
Yet Medium is now introducing a writing application built around AI.
This is not the same as rejecting AI entirely.
Medium describes its position as pro-human rather than anti-AI.
It recognizes the potential uses of AI technology while seeking to protect the value of human writing and creative work.
That philosophy also appears in TK's approach to privacy.
Medium stated that AI companies would not use writing stored in TK for model training without users' permission.
However, that statement should not be interpreted as meaning the service involves no external AI integrations or data processing of any kind.
The specific information processed to provide AI functionality, and the services involved, must be examined through the applicable terms and privacy policies.
In particular, connecting personal records to external AI tools may create additional forms of data access and sharing.
Medium's central commitment concerns preventing unauthorized use of users' writing for AI training.
That differs from prohibiting the use of AI technology itself.
In other words, Medium is seeking to incorporate AI while maintaining human creators' rights and control over their work.
11. Why Is MCP Important for Connecting Personal Writing to AI Agents?
TK also includes an MCP server capability.
MCP, or Model Context Protocol, is a standardized protocol for connecting AI applications with external tools and data sources.
In practical terms, it can allow AI systems to access information and records that users authorize them to use.
TK uses this approach to make accumulated writing and notes available as reference material for external AI agents.
Consider someone who has maintained extensive notes about a particular project over several months.
If an external AI agent can consult those records while assisting with the project, the user may not need to explain the entire background repeatedly.
Medium also said it provides a skill designed for integration with Claude.
This architecture could change the role of personal knowledge management tools.
Traditional note-taking applications primarily functioned as places where people stored information for later retrieval.
When connected through MCP, those records can also become sources of information that AI systems consult while performing tasks.
Personal writing and notes could therefore develop from archived material into reusable data assets supporting different AI projects.
However, this introduces security and privacy concerns.
Users need clear control over which materials external agents can access.
Permissions must prevent sensitive notes or unpublished manuscripts from being used in unintended tasks.
As personal knowledge becomes more useful to AI, maintaining control over that knowledge becomes increasingly important.
12. How Is TK Different from Notion or Obsidian?
Many of TK's features resemble those found in existing note-taking and knowledge management tools.
Obsidian is known for document linking and personal knowledge management.
Notion integrates documents, databases, and collaboration.
Evernote and Bear also provide recording, search, and document organization capabilities.
TK overlaps with these products in several respects.
Medium itself acknowledged drawing on experiences from existing applications during development.
However, TK emphasizes a different central objective.
Rather than focusing primarily on organizing notes, it aims to help users write more, develop their thinking, and eventually share their work with others.
To support that objective, Medium has connected notes, AI feedback, Social Libraries, and publishing into a continuous workflow.
Its ownership of an established publishing platform is an important distinction.
Writers using other note-taking tools may need to publish their work through separate services.
TK positions its connection with Medium as a central part of the product's value.
However, it cannot yet be assumed that TK offers better search, organization, or knowledge management performance than existing tools.
Independent comparisons are needed to evaluate functionality, reliability, and ease of use.
The currently visible differentiation lies less in individual technical features than in the product's attempt to connect personal records with writing, publishing, and relationships with readers.
13. Medium's New Competitors May Not Be Other Publishing Platforms
TK expands the competitive market in which Medium operates.
Traditionally, Medium has competed with blogs, newsletters, and online publishing platforms.
The main question was where writers should publish their work and build an audience.
TK moves into an earlier stage of the process.
It introduces competition over where writers record ideas, draft manuscripts, and collaborate with AI.
In this market, note-taking applications, document editors, AI chatbots, and knowledge management platforms can all become competitors.
This could influence Medium's long-term business strategy.
If writers accumulate their ideas and manuscripts in TK, the service may become part of their daily routines.
If they subsequently publish completed work on Medium, the two services could reinforce each other's value.
However, the arrangement does not necessarily mean writers will become permanently tied to Medium.
TK supports Markdown import and export.
Users can move their writing and records to other tools.
Nevertheless, TK's emphasis on cloud-based collaboration may involve different constraints from those of local-file-centered applications.
A native desktop application was also unavailable in the initial version.
Medium identified this as an area for future improvement.
TK's ability to retain writers over the long term will therefore depend on the actual user experience and the maturity of its features.
14. What Do a Free Plan and a $12 Monthly Subscription Mean for Medium?
TK offers both free and paid service options.
According to information available at launch, the free plan includes core features such as unlimited note creation, publishing, and collaboration.
The Pro plan was introduced at $12 per month.
Pro includes access to extended capabilities involving AI agents, skills, summaries, MCP, and related features.
This pricing structure suggests that TK could become more than an auxiliary feature of Medium.
It has the potential to develop into a service with an independent revenue model.
Medium's existing business has largely centered on published writing and the relationship between writers and readers.
TK could monetize activities that take place before publication.
Writers may be willing to pay for tools that organize information and provide AI feedback even before they publish a finished article.
However, it remains uncertain how many paying subscribers TK will attract or how the new business will interact with Medium's existing operations.
The information available at launch does not provide a sufficient basis for forecasting TK's revenue or profitability.
Medium said it had remained profitable since August 2024.
However, the announcement did not disclose detailed figures concerning TK's development costs, operating expenses, user base, or expected revenue.
The speed at which TK could become a sustainable independent business remains to be seen.
15. Will Human Writing Become More Valuable as AI-Generated Content Proliferates?
The most important claim in Medium's announcement concerns philosophy rather than technology.
Medium believes that the world already has an abundance of content generated by AI.
In that environment, building yet another tool that automatically produces articles may create limited additional value.
Instead, the company is focusing on the experience of humans thinking and recording their ideas.
This raises an important economic question.
As AI reduces the cost of producing text, the supply of written content may increase.
But an increase in the number of available articles is not necessarily the same as an increase in writing that readers find valuable.
Some readers want accurate information.
Others seek specialized expertise.
Still others value an author's distinctive perspective, personal experience, or ability to identify important questions.
As AI becomes increasingly capable of organizing general information, the relative importance of such differentiating characteristics may increase.
Medium views the thinking and experience accumulated by individuals as scarce resources in the AI era.
However, it would be inaccurate to assume that human-written content is always more valuable than AI-generated writing.
Human authors can make mistakes and exhibit bias.
AI-assisted writing may sometimes be more accurate or useful.
The important issue is therefore not simply whether a human or AI produced the sentences.
It also involves evaluating the evidence supporting an article, its originality, the responsibility taken for its claims, and the transparency of its creation process.
TK represents a decision to place human thinking and accumulated personal records at the center of that experience.
16. How Will Authors Be Evaluated in the AI Era?
Traditionally, authors have largely been evaluated through the finished work they produce.
What sentences did they write?
What arguments did they advance?
How persuasively did they express those arguments?
But as generative AI becomes more deeply integrated into writing, additional criteria may emerge.
What problem did the author identify?
Which sources did they select and evaluate?
Which AI suggestions did they accept or reject?
How much responsibility do they take for the final result?
These changes do not necessarily mean that the human author's role will disappear.
Instead, authorship may expand beyond producing sentences to include problem formulation, interpretation of evidence, judgment, and accountability.
Medium is attempting to incorporate this shift into TK.
Writers record their own ideas.
AI retrieves relevant information and provides feedback.
Earlier writing becomes material for new thinking.
When appropriate, authors may also disclose information about AI involvement and reference materials.
Whether this structure actually improves users' creative abilities requires further evaluation.
Nevertheless, the type of writing Medium wants to encourage is clear.
Rather than maximizing the volume of text produced by AI, it wants to increase the value of writing developed through human thought.
What Should a Writing Platform Sell in the AI Era?
Medium's introduction of TK illustrates a new challenge facing writing platforms in the age of generative AI.
AI can generate sentences quickly.
It can summarize research and suggest article structures.
For some tasks, it can produce drafts much faster than humans.
But writing involves more than generating a finished product.
It is also a process of organizing thoughts and discovering new questions.
It allows people to develop their arguments and judgments and communicate with others.
Medium is focusing on that process.
It has positioned AI not as a replacement for writing but as a tool that supports the development of human thought.
It enables users to accumulate earlier notes and manuscripts as personal knowledge and use them in subsequent writing.
It emphasizes feedback from actual people alongside feedback generated by AI.
It also introduces the possibility of disclosing AI involvement and reference materials associated with the creation process.
This approach offers one possible answer to where writing platforms can create value in the generative AI era.
Providing a place to publish completed articles may no longer be enough.
Services that support the formation and development of ideas could become increasingly important.
However, TK's success is not guaranteed.
It must demonstrate that it offers a better experience than existing note-taking tools and AI services.
The accuracy with which AI can understand and reuse personal records also requires evaluation.
Implementing meaningful authorship transparency and protecting users' control over their data will be equally important.
Nevertheless, Medium has raised a significant question.
In an era when AI can write faster than humans, can a writing platform give people a compelling reason to keep writing themselves?
Medium believes part of the answer lies in the accumulation of thought.
A note written today becomes an essay tomorrow.
Several essays develop into a distinctive perspective.
An accumulated body of thinking can become the foundation for a new project, research agenda, or book.
Throughout that process, AI can function not as a replacement for human thought but as a tool that connects earlier records and expands the possibilities for new ideas.
Even if AI can generate an almost unlimited number of sentences, the experience, judgment, and questions accumulated by one person over many years cannot simply be produced automatically.
That is the value Medium is attempting to capture through TK.
While much of the generative AI industry competes to make machines produce more writing, Medium has chosen to pursue a future in which humans think more and write more.
Ultimately, the success of this experiment will depend less on how convincingly TK can generate sentences and more on whether people using TK can think more deeply and produce more valuable writing in their own voices than they could before.