I was taking my daughter to nursery during her adaptation period and parked where we usually park on weekends. On weekends, parking there is free, so I had never really asked myself how you pay for it on a weekday.

It turns out the answer is: through an app.

To pay, the flow looks something like this:

Download the app → Verify email → Verify phone number → Add payment details → Add licence plate → Choose parking type → Find the exact parking zone → Pay

You can probably tell from the length of that flow that when you are taking your little daughter to nursery during her first days of adaptation, your mind is not exactly focused on configuring a parking app.

So I skipped the flow.

I parked, took my daughter inside, and left. Later, I got an €8 fine. And, strangely enough, part of me was almost grateful for how frictionless the parking experience had been in that moment.

I did not need to understand the system.

I did not need to register.

I did not need to configure anything.

I just parked and went where I needed to go.

The friction did not disappear.

It simply appeared later - as a fine and another half an hour spent figuring out how to pay it.

As someone who designs digital products, I kept thinking about this. We often treat automation as if it automatically removes friction.

But sometimes automation does something else:

it moves the work from the system to the user.

A person at the parking lot could have taken my payment. A barrier could have asked me for one simple transaction at exactly the right moment.

Instead, the automated system required me to understand how it works, provide context, configure myself as a user, make several decisions, and complete the right flow before I could perform a very simple task: park a car.

This raises a few questions that feel increasingly relevant far beyond parking.

1. When is onboarding actually necessary?

2. What is the bare minimum of data we need from a new user?

3. How much context should a user have to provide before the system can help them?

4. And when we automate a process, are we actually removing work - or just redistributing it?

This becomes especially important in AI products. We talk a lot about human in the loop in the context of control, oversight, approvals, and safety.

All of that matters.

But there is another role the human can play. Sometimes the human is there to absorb complexity.

They know how the system works.

They know what information matters.

They know which exception applies.

They know what to ask next.

And instead of making the user understand the entire system, they translate that complexity into a simple interaction. Remove that human, and the complexity does not necessarily disappear. Someone still has to deal with it.

Very often, that someone becomes the user. And this is where the question of delegation becomes more interesting.

If an AI product is supposed to take more work off the user’s plate, then simply removing people from a workflow is not enough.

The product has to absorb the work they were doing too.

Otherwise, we may call the experience automated while quietly increasing the user’s cognitive and operational load.

So perhaps one useful question for AI product teams is not only:

What can we automate?

But:

When we remove the human from the loop, who inherits the work that human used to do?

If the answer is “the user,” we may not have removed friction at all.

We may have just moved it.

Yesterday, my mom’s flight from Budapest to Larnaca was diverted to Paphos because of heavy rain, a thunderstorm and a waterspout.

She had already spent around 40 extra minutes in turbulence, not knowing where the plane would eventually land, so when she arrived in Paphos I told her not to wait for the airline to organise a transfer.

Just take a taxi to Limassol.

At that point I thought she had had enough uncertainty for one day. What I did not expect was that the taxi driver would take her without actually knowing where he was going.

He had no navigation and no Wi-Fi, so at some point my mom called us and handed him the phone. My husband explained the area where we live and sent a location pin to my mom’s phone, assuming the driver would simply open it and follow the route. He never opened it.

I suspect he did not know how to use navigation at all. At one point he was also calling us while driving on the highway, but that is probably a separate story.

The whole situation got to me much more than I expected.

Later, when everyone was finally home and nothing terrible had actually happened, I kept thinking about my own reaction. Ten or fifteen years ago, the same taxi ride probably would not have felt nearly as alarming.

Someone gets into a taxi. The driver does not know the exact address. The passenger calls. Someone explains how to get there. That used to be normal.

Today, I expect something very different. I expect to know who the driver is and where the car is. I expect to see the route, the estimated arrival time, whether the car is moving in the right direction. I expect to send a location pin and assume the other person can use it.

These things have become so ordinary that I barely think of them as technology anymore. They are just part of how the world works. Until they disappear.

And I think this is one of the more interesting parts of adaptation. We usually notice the moment a new technology enters our lives. We notice the first time we use navigation instead of a paper map, the first time a taxi appears on a screen, the first time we can see someone moving towards us in real time.

What we notice much less is the moment all of this becomes expected. At some point, it stops feeling like an improvement. It becomes the baseline.

And once that happens, the old way of doing things can suddenly feel much harder than it ever did before. That is probably why yesterday felt like travelling 15 years back in time.

The situation itself was familiar in an old-fashioned way. What had changed was me. My expectations had already adapted to a different environment.

I spend a lot of time thinking about this in the context of products and AI. Right now, a lot of the conversation is still about adoption.

• Who is using AI?

• Which products have added it?

• Where can a company introduce an AI feature?

But I think the more important shift will happen through expectations. People are getting used to software that can remember context, draft something instantly, summarise information, or remove parts of a workflow that used to be manual.

Once those experiences become common enough, they start changing how everything else feels. A product that repeatedly asks for information it could already know begins to feel unnecessarily demanding.

A workflow that requires five screens can start to feel strange once people get used to expressing intent directly.

Waiting days for a first draft feels different once creating one takes seconds.

Manual repetitive work feels more visible when we know a machine could probably take part of it away.

None of these changes happen because a user suddenly decides that every product needs AI. Expectations shift more quietly than that. People simply get used to a new level of convenience, context, speed or assistance somewhere else, and then carry that expectation with them.

That is why I keep coming back to the distinction between AI adoption and AI adaptation.

Adding AI changes the product. Adaptation starts when the environment around the product has changed enough that the old assumptions no longer hold.

Yesterday, a taxi ride made that idea unexpectedly tangible for me.

Ten or fifteen years ago, not knowing exactly where a taxi was would have been completely normal. Yesterday, the absence of that information felt deeply uncomfortable.

The world around me had changed gradually enough that I had stopped noticing what I now considered normal.

And I suspect AI will do the same thing to many more expectations over the next few years.

The most interesting question may not be what new things AI will allow us to do.

It may be what we will soon find surprisingly difficult to live without.

In this environment, we meet a character who will go through almost everything the arrival of AI has brought us. The challenges, of course. But also the possibilities. Because in what other world can you suddenly fold up like a telescope and find yourself thinking:

I wish there were instructions for people who fold up like telescopes.

There are no instructions, of course.

But that does not stop Alice. Perhaps she is still too young and naive to fully understand which things are generally considered impossible. Or perhaps she possesses a rare ability:

When Alice finds a bottle labelled DRINK ME, she first checks whether the word “poison” appears anywhere on it. That is sensible. After a brief investigation, she takes a sip. And the world immediately becomes a little bigger. Or Alice becomes smaller. In Wonderland, those are almost the same thing.

Suddenly, she can pass through a door that seemed unreachable only moments ago. On another occasion, she grows so large that she begins to worry about her own feet. I do not think Alice was prepared for that. What saved her was curiosity.

The almost childlike curiosity that woke up in a great many perfectly grown adults when AI became widely available. Everyone received a bottle. It did not say DRINK ME. It said:

TRY ME.

And regardless of whether you were seven, thirty-seven, or slightly older, there was suddenly an opportunity to fall into a new world and see what was waiting at the bottom.

There, a single sentence can produce an image. A question returns with an answer from a conversational partner who sounds as though it has read everything ever written, including books that have never existed.

Work that once took a week can sometimes be compressed into a few minutes. Curiously, after a while, even more work finds its way into the space that has been freed.

In Wonderland, changing size never solved a problem completely either. When Alice became small, she could no longer reach the key on the table. After the next transformation, the key was close enough, but the door was no longer suitable for a person of her new dimensions.

That is how an environment behaves when it has not yet acquired an instruction manual. The ability to create ten options in a minute sounds wonderful until someone has to decide which of them makes sense.

Eventually, Alice manages to return to approximately the right size.

Not long after, she meets the Caterpillar.

Who are you?

he asks.

It should be an easy question. But Alice has already changed size so many times that she is no longer entirely certain.

I - I hardly know, sir, just at present,

she says.

This too feels familiar. The names of our roles may remain the same while the work inside them quietly changes shape.

• A writer can generate without writing every sentence.

• A designer can produce a working prototype without designing every screen.

• A manager may find themselves coordinating not only people, but agents.

The title remains in place. The boundaries underneath it begin to move. Perhaps the hardest question in a new environment is not what you can now do. It is who you become once you can do it.

Share

Somewhere after the pool of tears - the same one in which it is easy to find yourself when the new environment suddenly begins discussing whether it still has a place for you - Alice meets the Cheshire Cat.

He is ready to discuss roads, directions, and possible destinations even though Alice has not yet decided where she wants to go. In this respect, he is rather like a modern AI assistant. Both are happy to continue the conversation. Both can sound remarkably convincing. And neither suffers greatly when the original question disappears somewhere along the way.

Next comes a tea party that never ends. Everyone talks, changes the subject, and moves around the table. The conversation appears extraordinarily productive. It does not, however, lead to a result. There is something strangely familiar about this…

Later, Alice finds herself in court. The Knave of Hearts is accused of stealing some tarts. The jurors write down their own names so they will not forget them before the trial is over. The witnesses answer questions that have very little to do with the case. A letter without a signature, recipient, or comprehensible meaning is presented as evidence.

The King searches it for significance. And, naturally, he finds some. When a text is sufficiently unclear, almost anything can be discovered inside it. Eventually, the most efficient possible sequence is proposed: the sentence first, the jury’s decision afterwards.

At this point, Wonderland stops feeling quite so distant. AI produces a convincing answer. The explanation can be requested later. Only afterwards does someone check whether sufficient grounds existed. Sometimes they did. Sometimes there is an unsigned letter in which everyone would very much like to find evidence. And sometimes the confidence of the answer begins to serve as evidence by itself.

By now, however, Alice is no longer the same small girl who fell down the rabbit hole. During the trial, she begins to grow again. At first, almost imperceptibly. Then she becomes large enough to stop being afraid of the cards. It's not the rules that became more logical. Alice has changed.

She has altered her size several times, become lost in conversations without conclusions, and watched an incomprehensible text transform into evidence simply because someone wanted badly enough to find evidence inside it. Statements that once sounded like the unquestionable laws of the new world no longer seem quite so absolute.

Alice looks at the court and finally says:

You’re nothing but a pack of cards.

Lewis Carroll could not have known about AI. But he certainly knew something about a world in which possibilities appear without instructions, scale changes, conversations continue without decisions, and confidence attempts to replace evidence. He knew something else too.

Leave a comment

What saved Alice was not learning every one of them. That was probably impossible anyway. What saved her was curiosity. She experimented, observed the consequences, argued with the rules, and occasionally admitted that she understood nothing at all.

Gradually, one thing became clear: the world had not lost its logic. The logic simply belonged to the environment, not to the person who had just entered it.

The bottle labelled TRY ME is still standing on the table. No one knows whether its contents will make a person larger, smaller, or simply produce several more options for a decision that still has not been made. But perhaps size is not the important part.

And one day becoming large enough to look at its loudest rules and say:

You’re nothing but a pack of cards.

In Wonderland, a sentence could be passed before the evidence was considered. Our new environment has a sentence of its own - one whose meaning depends entirely on where the pause is placed.

Adapt not. Optimise.

Do not change. Simply learn to perform the same work faster and more efficiently.

Or:

Adapt. Not optimise.

First, understand which world you have entered. Only then decide what is worth doing faster. The words remain the same. The sentence is entirely different. So the question is not whether a choice must be made. The question is where you will place the full stop.

The realization was painful because the person I had been before becoming a mother still lived inside my plans. She planned full working days. She could spend hours refining a design, exploring alternatives, going back to an idea the next morning, and giving it another try.

My environment had changed. My expectations had not. For a while, I treated this as a productivity problem. Maybe I needed a better schedule. A more precise system. More discipline. Better tools.

Perhaps AI could help me I thought. But I quickly realized that even with AI, I could not simply continue doing the same work faster. Two hours were still two hours. I could not optimize my way back into my previous life.

I had to adapt.

The old strategy belonged to the old environment. This was not the first time my environment had changed. I had moved countries twice and, each time, rebuilt a significant part of my life. I think adaptation became one of my skills before I even had a name for it.

But this time, I was not adapting to a new country. I was adapting to a new amount of time. And time changed the value of everything. Tasks that once felt normal became too expensive. Work that required long, uninterrupted concentration became difficult to sustain.

The question was no longer:

How can I do everything I used to do?

It became:

What is still worth doing in this environment?

That is a very different question.

And I think it is the question many of us should be asking about AI. AI is not only changing how quickly we work. Most conversations about adapting to AI begin with tools.

• Which tool should I learn?

• How do I write better prompts?

• How can I automate more of my workflow?

These questions are useful. But they assume that the old result is still the right result. When an environment changes, this assumption becomes dangerous.

My problem was not that I had become too slow at living my old life. My old life was no longer available. Similarly, AI is not only helping us complete familiar tasks more quickly. It is changing the environment in which those tasks used to have value.

Using AI to accelerate an old workflow may help for a while. But at some point, we have to ask whether we are optimizing for an environment that no longer exists.

Once I accepted that I could not fit my previous working life into two hours, I started asking myself what I could build around instead. What was my strongest skills, professional superpower?

What could still create value without requiring an endless amount of time? Several answers came from my earlier work in sales and product design.

I had learned to ask good questions.

I had learned to listen carefully.

And I had learned to observe.

I chose observation.

This skill was the one I could still practice in the life I had. I could observe while working with a client. While testing an AI tool. While watching a team make a decision. While realizing that AI was not simply changing products. It was changing roles, relationships, and the structure of teams.

I did not need a full day to notice something. I needed enough attention to recognize that it mattered. So I started writing these observations down.

Eventually, some of those observations became posts. Some became articles. Some started turning into frameworks. And some are slowly becoming chapters of a book.

I started writing because my environment changed, and observation was the ability I could build a new way of working around.

Adaptation begins with recognition.

We often imagine adaptation as invention. A dramatic transformation. A completely new career, identity, product, or strategy.

But adaptation often begins much earlier and much more quietly. It begins when you recognize that something has changed.

Before creating a new strategy, we have to see the new environment clearly. This is difficult because old environments remain alive inside us. We keep trying to become better at a game whose rules have already changed.

My two productive hours made this impossible. The constraint was too visible to ignore. AI creates a less obvious constraint.

What to do when the environment changes

I do not think adaptation begins with immediately learning every new tool. This looks more like optimisation to me.

1. Name what actually changed

“AI is changing everything”

Is too broad to be useful. What changed in your work specifically.

• Can people now produce a first draft without you?

• Can stakeholders generate their own prototypes?

• Can a developer make interface decisions that previously required a designer?

• Can customers compare more options, move faster, or expect instant support?

• Can work that once took a week now be completed in a day?

The more precisely you describe the change, the easier it becomes to see its consequences.

2. Notice which old strategy is becoming too expensive

When my time changed, certain ways of working became unsustainable. The same is true when technology changes.

• Which activities still consume most of your time but create less value than before?

• Where are you polishing output that can now be generated cheaply?

• Where are you protecting a process simply because it once justified your role?

• Where are you trying to preserve the previous shape of your work instead of responding to what the work now requires?

Adaptation often involves stopping something before starting something new.

3. Find the ability that still compounds

When execution becomes cheaper, other abilities become more visible.

Judgment.

Taste.

Context.

Responsibility.

• The ability to frame the right problem.

• The ability to recognize which generated option should not exist.

• The ability to notice patterns across products, teams, and systems.

• The ability to ask a question that changes the direction of the work.

These abilities are not always easy to demonstrate in a portfolio or measure in a productivity dashboard. But they compound with experience.

The important question may not be:

What can I do that AI cannot do?

That line will keep moving.

A better question is:

What ability becomes more valuable because AI can do more?

4. Build around the new constraint

For me, the constraint was time. For someone else, it may be a role that is becoming less defined, a market that is becoming more crowded, or clients who no longer want to pay for execution alone.

The instinct is often to resist the constraint. To prove that the old model can still work. But constraints are information. My two hours forced me to stop building my work around the assumption that I had eight.

AI may force us to stop building our roles around the assumption that execution is scarce. Once you accept the new constraint, you can design around it.

5. Observe before you optimize

AI makes action incredibly easy. We can generate the deck, the interface, the strategy, the article, the research summary, and ten variations of each.

The temptation is to accelerate immediately. But speed only helps when the direction still makes sense. Before optimizing a workflow, observe what is happening around it.

• Which decisions are changing?

• Where are new problems appearing?

• What are people now responsible for?

• What is no longer being discussed because everyone is too busy generating?

Adaptation does not begin with producing more. It begins with seeing better.

Your environment may have already changed

I sometimes wonder whether I would have started writing if I had still had a full uninterrupted working day. Perhaps I would have continued designing more. Doing the things I already knew how to do.

The constraint removed that option. It forced me to look for another way to create value. AI is beginning to do something similar to many professions.

The old strategies may still work for a while. Sometimes they may even work better with AI. But adaptation is about noticing what the new environment rewards.

My two productive hours taught me that the answer is not always to become faster at what you used to do. Sometimes the answer is to discover what you can do now that still matters.

For me, it was observation.

Then writing.

I am still finding out what comes next.

Every day, I write down one small thing I notice while working with AI. Not predictions about what AI will do in five years. Usually, they are much smaller moments:

• My editing agent refusing to simply edit.

• A stakeholder building a prototype.

• An AI tool changing more than I asked it to change.

• A writing pattern I suddenly began seeing everywhere.

Separately, they looked like observations about different tools and situations. Together, they pointed to the same shift:

A role is the position someone holds in a team. A skill is what they can do. A responsibility is the outcome, perspective, or decision they are expected to protect.

For a long time, these three things were bundled together. A designer created prototypes and protected the user’s perspective. A developer wrote code and protected the technical integrity of the system. A copywriter shaped an argument and owned what it meant.

AI is beginning to unbundle this package. A stakeholder can now create a prototype. A designer can use AI to generate code. An AI agent can edit an argument - and even challenge it. Almost anyone can produce polished writing.

Skills - or at least access to the outputs those skills once produced - can now move beyond the roles they traditionally belonged to. But responsibility does not automatically move with them.

A stakeholder may generate a convincing interface without becoming responsible for representing the user throughout the project. A designer may produce code without becoming responsible for its long-term technical integrity. An AI agent may rewrite an argument, but it will not be responsible for what that argument means, whether it is true, or whether it should be published.

This is where a new gap appears.

I think of it as the responsibility gap.

The skill moves. The responsibility stays behind. And when this new distribution is not made explicit, capabilities spread faster than accountability, judgment, and ownership.

This week, I noticed five signs of that gap.

Each one suggests something practical we may need to change in the way we work.

1. Everyone can prototype - but the user’s perspective can become ownerless

Developers can design. Designers can code. Stakeholders can generate working interfaces before the team has even agreed on the problem. The traditional boundaries between roles are becoming less visible.

But the perspectives attached to those roles do not automatically become shared. When everyone can produce a solution, solution production is no longer the scarce skill.

Insight: Assign ownership of perspectives, not only deliverables

Before starting an AI-assisted project, explicitly assign who is responsible for protecting:

• the user’s perspective;

• business viability;

• technical integrity;

• risk and compliance;

• the final product decision.

This does not mean that only one person may contribute to each area. It means that someone remains accountable for making sure the perspective is not lost. A useful question to add to a project kickoff is:

Without explicit ownership, responsibilities do not necessarily become shared. They may simply become ownerless.

2. AI makes generation cheaper, but creates invisible supervision work

AI prototyping feels fast because the first output appears almost instantly. But the speed of generation hides another category of work:

• reviewing,

• correcting,

• restoring,

• constraining,

• and explaining what should not have changed.

The visible task becomes cheaper. The invisible supervision around it grows. The human does not disappear from the process. The human moves from creator to reviewer, controller, and repair layer.

Insight: Measure correction cost, not only generation speed

Do not evaluate an AI prototyping tool only by how quickly it produces the first version.

Also track:

• how many unintended changes appear per iteration;

• how often previous decisions need to be restored;

• how much time is spent reviewing the output;

• how often context must be repeated;

• how many iterations are required before the result is safe to use.

One simple metric could be:

If AI generates something in two minutes but requires forty minutes of cleanup, the meaningful productivity story is not the two minutes. Generation is only one part of the cost.

3. AI can improve the sentence - and challenge the author

I write instructions to my editing agent in Ukrainian because it feels more natural to me. Its job is simple: improve my English comments.

This week, I wrote:

AI would never come up with that detail on its own.

Instead of polishing the sentence, the agent stopped me. It pointed out that AI sometimes does suggest details like that. What it cannot do is actually live through the moment behind them.

I was not expecting my editing agent to defend AI. But that was not really what happened. It challenged an assumption that weakened the point I was trying to make. The result was probably more accurate.

It also changed the relationship. The tool was no longer only improving how I expressed the thought. It was participating in the thought itself.

Insight: Separate editing from decision-making

When using AI to edit or refine important work, ask it to distinguish between different types of intervention:

• language correction;

• structural improvement;

• factual challenge;

• change of argument;

• added assumption.

These are not the same thing.

A useful prompt could be:

Improve the writing, but flag separately any place where you disagree with the premise, change the argument, or introduce a new claim.

This keeps AI’s contribution visible. The final question should not only be:

Is this better written?

It should also be:

Is this still my argument?

4. AI can improve our writing - and make all of us sound the same

This week, I learned the word anaphora. It is a writing technique that repeats the same word or phrase at the beginning of consecutive sentences. To be honest, I had never heard the term before.

AI clearly had.

Once I learned the name, I started recognizing the pattern everywhere:

The same beginning.
The same rhythm.
The same spacing.

One sentence. A new line. Some space. Another sentence. AI did not invent anaphora. Human writers have used it for centuries. But AI can identify effective writing patterns and reproduce them at extraordinary scale.

That is why AI-assisted writing can sound polished and strangely familiar at the same time. The risk is not that the pattern is bad. The risk is that a useful technique becomes an automatic default.

When many people use similar tools trained on similar writing patterns, “better writing” can begin to converge into the same voice.

Insight: Review AI writing for patterns, not only errors

After asking AI to improve a text, do a separate pattern review.

Look for:

• repeated sentence structures;

• excessive one-line paragraphs;

• artificial contrasts;

• sets of three;

• dramatic conclusions;

• phrases you would not naturally use;

• repetition added mainly to create rhythm.

A useful follow-up prompt is:

Identify the rhetorical devices and recurring structures you introduced. Do not rewrite the text.

Then decide which patterns genuinely serve the idea and which only make the text sound generically polished. AI should help us use writing techniques intentionally. It should not choose our voice by default.

5. Context is becoming something we need to design

Sometimes I add a thought to ChatGPT as a note. And sometimes ChatGPT interprets it as a question.

This time, I wrote:

Maybe the branch feature is ChatGPT’s attempt to give context a structure.

ChatGPT answered that branching provides context through controlled inheritance rather than by simply adding more of it. So, assuming we trust ChatGPT’s competence on the subject of ChatGPT, we have confirmation. Or at least a very confident explanation.

But the observation itself points to something larger.

As AI conversations become longer, context stops being an invisible technical mechanism. Users need to decide:

• Which assumptions should continue?

• Which decision belongs only to this experiment?

• What should a new branch inherit?

• How can we return to a stable point?

More context is not always better. Sometimes more context simply means more assumptions, more contamination, and more opportunities for the system to reinterpret something that had already been decided.

Insight: Treat context as part of your information architecture

For recurring AI work, do not keep everything in one endless conversation. Create explicit context boundaries:

• one thread for exploration;

• one for approved decisions;

• one for execution;

• one for experiments that should not affect the main direction.

Before continuing a long AI interaction, state:

In AI-native products, context management may become as important as navigation. Users will need to understand not only where they are, but also:

• what the system remembers,

• what it inherited,

• what it considers a decision,

• and what can still be changed.

Context is becoming something we need to design.

Five practices I am taking into next week

These observations led me to five practical rules.

1. Assign ownership of perspectives

When role boundaries blur, explicitly decide who protects the user, the business, the system, and the final decision.

2. Include correction cost in productivity

Measure how much human supervision follows the instant output.

3. Ask AI to label the type of intervention it makes

Editing, challenging, and introducing a new assumption should not look identical.

4. Audit AI-generated patterns

Review not only whether the output is correct, but whether it still reflects your reasoning and voice.

5. Design context boundaries

Decide what AI should inherit, what remains experimental, and which decisions are fixed.

The larger mutation

The common thread across all five observations is ownership.

• Who owns the user’s perspective when everyone can build?

• Who owns the unintended changes inside an AI-generated prototype?

• Who owns the argument after AI improves it?

• Who owns the voice when the writing pattern came from the model?

• And who decides which context should survive?

That part still has to be designed. The practical task is not to force skills back into their old roles. It is to make the new distribution of roles, skills, and responsibilities explicit - and to reassign the perspective, judgment, and accountability that used to travel with those skills.

Otherwise, skills will continue to move.

And responsibility will continue to stay behind.

Throughout the week, I publish short observations about how AI is changing products, roles, and the way we work. Every week, I connect all of them into one practical digest.

Subscribe to receive the next Observation Digest.

Subscribe now

Imagine standing in a tropical forest millions of years ago. You’re hungry. Somewhere above you, hidden among the leaves, hangs a ripe red palm fruit. You know it’s there. You move closer, scanning the branches, trying to catch a glimpse of it before climbing the tree.

But stop. Let’s pause for a moment.

That isn’t actually what our ancestors saw. Evolutionary biologists believe early primates hadn’t yet developed the color vision most of us take for granted today. The fruit wasn’t red. Not to them.

The forest was a world of greens, yellows, light, shadow, and contrast. Finding food wasn’t simply a matter of looking in the right direction. It meant searching branch by branch, relying on shape, texture, smell, memory, and sometimes pure luck.

Then, over many generations, something remarkably small happened. A genetic mutation produced an additional photopigment receptor. For the first time, reds and greens became easier to distinguish. The fruit became easier to find. The forest hadn’t changed. What changed was the ability to perceive it.

That story stayed with me. Not because I was particularly interested in biology. Quite the opposite. Biology was never really my thing. Which is why I still find it slightly ironic that it ended up giving me the language I’d been searching for.

For months I kept noticing the same feeling during product work. Conversations had changed. Discovery had changed. Even the relationship between people and software felt different.

Yet every word I reached for felt slightly wrong.

• Transformation.

• Automation.

• Disruption.

• Evolution.

Each one described part of what I was observing. None of them described the whole picture. I wasn’t looking for a better metaphor. I was looking for better language. Then I remembered another question that had fascinated me for years.

For centuries, scholars have debated why Homer described the sea as wine-dark instead of blue.

Was it language? Was it perception? Or was it simply a different way of experiencing the world? No one knows for certain. But the question itself stayed with me. How much of reality depends not only on what’s there…

…but on what we’re able to notice?

That’s when I realized I wasn’t searching for a better metaphor. I was searching for a better name.

Mutation.

Because, in biology, mutation is rarely the end of the story. It’s the beginning. A small change that quietly reshapes what becomes possible. That was the language I’d been looking for.

Not a metaphor.

A language.

A way to describe many of the changes I was beginning to observe across products, organizations, and the people building them. I eventually started calling it The Mutation Era. Biology had already developed a language for describing something I was struggling to name.

That idea eventually became the starting point for a book I’m writing called The Mutation Era. It’s an exploration of what happens when the environment changes - and everything inside it begins to adapt.