In the era of artificial intelligence, how to reimagine your work

God Translation Bureau is a compilation team under 36Kr, focusing on fields such as technology, business, workplace, and life, and mainly introducing new technologies, new ideas, and new trends from abroad.

Editor's note: In the era of rampant AI output, content stacking is dead. Disassembling ideas like Figma did with files is the irreplaceable "native work". This article is from a compilation.

I haven't spoken here for a long time.

At the end of last year, I started to ponder a question - since "producing" work results is no longer the bottleneck, what form should work take now?

To put it more broadly, what is "AI-native" work? In an era when producing results is no longer difficult, how should we, those who have built our careers around "producing high-difficulty results" in the past, reshape our work?

But before delving into these, I want to introduce something I've been working on for the past few weeks.

The Dynamic Companion Guide to "Reshuffle"

Many of you have enjoyed reading my book "Reshuffle".

I'm very pleased to announce that the companion guide to "Reshuffle" is officially launched.

I mentioned in the book that I would launch the companion guide to "Reshuffle" in early 2026.

When I first started preparing, I originally planned to write it as another book. But as I delved deeper into the work, I realized that the book format might not be the most suitable for what I wanted to do.

Books are still an extraordinary medium for spreading grand ideas in a persuasive and structured way. "Reshuffle" itself was written with this purpose in mind. But in the era of artificial intelligence, many ideas need to be reached in different ways. Some need to be explored, queried, reorganized, navigated, and applied. They require interaction, not just exposition.

I believe that books need to be disassembled and recombined.

This realization made me reevaluate this companion guide. Instead of making it a static book, I launched it as an interactive product - a dynamic tool that accompanies "Reshuffle".

Currently, it contains three core sections, and its content will continue to expand over time.

First and foremost, you can use the mind map of "Reshuffle" to sort out the ideas in the book. This map clearly shows the core concepts and their internal relationships.

You can explore specific main lines or paths in the map. For example, the "coordination" main line:

Or, you can also study along one of the four paths in the book - these paths explore the impacts brought by artificial intelligence at four levels: positions, organizations, ecosystems, and the broader macroeconomy.

Finally, if you just want to browse around, you can follow the "scrolly-telling" plot on the home page.

AI-Native Books/Newsletters

The Disassembly and Recombination of Books/Newsletters

This brings us back to a core question:

In the era of artificial intelligence, how should we reshape our work?

For most of the time in the knowledge economy, we have evaluated knowledge-based work through output. We look at reports, tables, program codes, design drafts, presentations, books, or newsletters. These tangible products have become the visible proof of work results.

But this has caused a subtle distortion. We began to overvalue these products themselves while underestimating the problems these products were supposed to solve. For example, the value of a book lies in its ability to help readers understand, believe, remember, apply, or spread a certain idea. The value of a newsletter lies in its ability to create rhythm, attention, and connection around a series of ideas.

However, these products themselves may not be the best solutions to the problems they were originally intended to solve.

Artificial intelligence has revealed this difference because it has made product production easier and reduced the resistance significantly. When output becomes rampant, a well-packaged product is no longer so rare and precious.

What is truly valuable is to figure out what problems really need to be solved and how to present ideas in the best form to accomplish this task beautifully.

This transformation has changed the nature of knowledge-based work. Value is no longer reflected in producing a single finished product but in designing a system - through this system, ideas can be structured, reorganized, and presented in different forms for different purposes. The same core idea may need to evolve into a chapter of a book, a mind map, a visual chart, a decision-making tool, or an interactive companion application.

Therefore, the real work is no longer to produce specific products. More importantly, it is the "idea architecture". It requires us to understand what role an idea needs to play, disassemble it into reusable components, and establish corresponding mechanisms to recombine these components into the most suitable output for specific scenarios.

In a world where output can be generated at any time according to demand, the scarce skill is no longer just "producing output". It is to understand what this output is for and master the logic of disassembly and recombination in this process.

Take the companion guide to "Reshuffle" as an example. It is by no means a perfect ultimate solution, but it does two things that you can see at a glance.

It disassembles a book into core idea components.

Then, it allows these ideas to be reborn in new ways through different recombination forms.

The mind map is an example of recombination.

The study path is another example.

Once you browse these pages, you will find that they are just different packages of the same basic components.

In this regard, the book itself is also a specific combination of these ideas, which are strung together through a specific narrative style to bring about a certain cognitive transformation for readers.

However, once readers finish reading the book, it may no longer be the best product to meet their next needs - such as querying ideas, applying ideas, or expanding on them.

And this is where building a reusable "idea architecture" by disassembling and recombining books comes in handy.

Reimagining the "Work Unit" in the Era of Artificial Intelligence

What does it really mean to do "AI-native" work?

Most people answer this question by simply regarding artificial intelligence as a tool. They use it to summarize a book, turn a keynote speech into a slide gallery on social media, turn the gallery into a long tweet, turn the tweet into a podcast, and then turn the podcast back into a blog post.

They are faster, cover more channels, and feel extremely productive.

But this is the same as what Adobe did during the cloud computing wave: the same content, just distributed through new channels.

This is just "repurposing". It seems like a transformation, but it doesn't change any underlying structure. It is still built on the logic of "outputs", just like Adobe used to be built on the logic of "files".

Figma didn't simply move design files to the cloud like Adobe did. Figma completely eliminated the file logic. As the article says:

It replaced files with "elements" - such as buttons, icons, or font styles - as the basic unit of work.

Thanks to this element-based architecture, Figma users can create a shared library of reusable design components (such as buttons, icons, font styles, and color palettes), which teams can use across multiple files and projects. Designers don't need to copy these elements in each file but only need to reference the same "single source of truth".

At this time, "files" have become a specific recombination form of these elements, rather than isolated and fragmented objects.

Modifications and permissions can be tracked and managed at the level of design elements. Each element is addressable in the database: when you modify a component once, the modification will be automatically synchronized to all places where it appears.

This ensures consistency, simplifies the update process (one modification, all updates), and enables cross-functional teams to work together under a unified visual standard. The shared library has pushed design work from isolated file ownership to coordinated system-level collaboration.

By changing the work unit from "files" to "elements", Figma has achieved real-time collaboration, created a shared design environment with a lower participation threshold, and made Adobe's model increasingly limited by its own architecture.

Figma's move and Adobe's move seem similar from afar but are actually diametrically opposed. One is to push the same rigid products to new channels; the other is to disassemble the products and let new products be born from a brand-new recombination.

Adobe's action is a story about "distribution", while Figma's action is a story about "architecture".

Why the Written Form Has Always Been a Compromise

Every result we produce - whether it's a book, a report, a PPT, or a course - is a package. Inside any package, there are: frameworks, arguments, cases, evidence, rebuttals, definitions, stories, and decisions. And what holds this package together is the narrative - a single order of arrangement selected for these ideas.

However, narrative is just one of the many recombination logics aimed at solving specific problems.

The ideas in your mind are associative in a network. When you think of a concept, a dozen related concepts will immediately come to mind - examples, special cases, balancing factors, and practical applications. No writing technology can directly convey this network structure. Therefore, the author has to compress the intricate associated ideas into a selected linear context, discarding all other possible arrangements that might also work because there is no place for the rest. Then, the reader decompresses and reconstructs these connections in their own mind based on the only path they get.

To put it bluntly, a book is just a "lossy" expression between the author's understanding and the reader's reconstruction.

Once you can see that this linear result (whether it's a book, a table, or a product design document) is actually an attribute of the medium rather than an attribute of the idea itself, you will find the path from the Adobe model to the Figma model.

Disassemble First, Then Recombine

This involves two completely different steps.

"Disassembling" means realizing that the final product is not the basic unit. You no longer regard a book (or any product/result) as a whole that needs to be chopped into various formats, but start to view the ideas in the book as individuals with independent vitality. You no longer ask "How can I squeeze more value out of this book?" but start to ask "What are the core ideas I really have, and what can they become when they exist independently?"

This is by no means as boring as "content repurposing", which is what Adobe does - still regarding the final output as the work unit and just repackaging it for new channels.

And "recombining" is a miracle that becomes possible only after disassembly: it can produce combinations that the original linear narrative can never express. It creates something brand-new for new scenarios, new audiences, and new problems. At this time, the way each idea is expressed becomes a function of its context.

"Repurposing" keeps the original package intact and only changes the wrapping paper. "Disassembly and recombination" changes the package itself - once you master this ability, you can recombine it again and again for new problems that the original output never anticipated.

This is not the reuse of products but the reuse of the "internal connections between ideas". What you reuse is no longer the rigid output but the connection between ideas, and yesterday's output is just one way to package these connections.

Injecting Superpower into the Creative Economy

This has a huge and far-reaching impact, far beyond what you can do with a book, a newsletter, or any product you create.

Let's take books as an example again.

As early as 2012, when I first started taking my research and writing seriously, I recorded almost every idea I encountered. This was largely for my own research habits - to put everything in one place to find connections between seemingly unrelated ideas.

For most of the past time, this process was completely manual. I used Workflowy as my core knowledge base. Although there were more complex mapping tools at that time, I found that most of them were too cumbersome for daily use. Workflowy was very useful because its visual and usage experience was very similar to Microsoft Word, while at the same time giving me the flexibility to move large chunks of ideas to new hierarchical structures at will.

In the past nearly 15 years, calculating an average of 300 days a year and an average of 25 to 30 new ideas added per day, a huge corpus containing more than 100,000 ideas has been accumulated here. I say "more than 100,000" because on days when I was really immersed in a certain topic, I could easily add 100 to 150 idea components.

However, only a very small part of all this corpus has finally become books and newsletters.

More importantly, my ability to find combinations and causal relationships between these ideas is limited, not only because of my limited cognitive bandwidth but also because as I add more and more ideas, the manual work required for reclassification and hierarchical reconstruction is increasing exponentially.

Here is how I managed it in the past 15 years:

Every day, I would put the content I read and related links into Workflowy. From time to time, I would go back and clean it up. In December every year, I would take the last two weeks to review the notes of that year, reorganize them, and integrate them with the ideas collected in previous years. In fact, I usually stayed at a seaside resort at that time, with nothing to do but immerse myself in all kinds of ideas from the past year, which made it an annual ritual I really looked forward to.

That system has helped me a lot. It has helped me write three books. More importantly, it has allowed me to continuously establish weak connections between everything I read, learn, and think about.

But I now realize that a knowledge base like this only has real value when it is "activated". If it just sits there as a huge archive, its usefulness is quite limited.

And this is where generative artificial intelligence (GenAI) becomes particularly interesting.

Because it can understand language, recognize concepts, and establish connections between them, it can help transform a largely implicit work system into something more modular, searchable, and generative.

Over the years, every article, essay, newsletter, and book chapter I've written has essentially been an attempt to string together the ideas that already exist in my Workflowy in a narrative way. Like most knowledge workers, most of this work has been done implicitly in the subconscious. But with generative artificial intelligence, there is now an opportunity to make those underlying concepts, connections, and patterns more explicit.

In many ways, this is an extension of the question I've been thinking about since I started writing "Reshuffle": the componentization of knowledge, the modularization of ideas, and the sharp decline in the conversion cost between previously isolated fields.

For a long time, I've been deeply interested in how these transformations apply not only to industries and organizations but also to the world of ideas, concepts, and theories.

In the past few weeks/months, I've been trying out various tools, mainly Claude Code, as well as some tag and classification management systems, to organize and reconstruct my knowledge base. The greater ambition is not just to create a better archive but to build a living system that can recognize associations, recombine ideas, and help me navigate through all the in-depth content I've collected over the past decade.

In this process, I want to continue to repackage these concepts into brand-new combinations and offer them to the readers and consumers of these ideas.

Narrative forms such as books and newsletters will continue to play a role where they can generate the most value. However, once the idea architecture is built, there is still a lot that can be done.

Next Steps!

If you find this approach appealing, I'd love to hear how you've applied (or plan to apply) it in your work.

Currently, the next "recombination" I'm working on is a job index - a product that obtains market data and applies the arguments in "Reshuffle" to determine the degree to which a specific job is reshuffled.

Translator: boxi.