I Was Tired of Living in Two Different Worlds

Language Beyond Binary (LBB) began in 2023 as a weird little interactive AI notebook because, quite frankly, I was tired of the machine. Red versus blue. Black versus white. Us versus them.

Three years, several names, one Little Blue Butterfly, an unreasonable number of questions, and a truly absurd amount of language later, I finally built the thing.

Or at least the first real version of it.

I should probably begin with language, because that is where almost everything begins for me.

I love language. I mean, I LOVE language. Probably too much.

Life without color will do that to you.

I was born with achromatopsia. I do not experience color the way most people do, so words have always helped me reach distinctions my eyes cannot. Someone says hunter green. Sky blue. Golden hour. Hazel eyes. I do not suddenly see what they see, but something crosses.

Describe it to me. Tell me what you see. Give me another comparison. Move the angle.

Language can do something extraordinary: it can let one human being borrow, imperfectly, another human being’s view.

I became a lawyer.

Of course I did.

But Language Beyond Binary did not begin as a legal project, an AI-safety framework, a philosophy of intelligence, or educational software. I was just increasingly irritated by the feeling that people living in the same country could open their phones and encounter radically different versions of reality.

I would look at my newsfeed and then look at my father-in-law’s. Different stories were salient. Different harms were vivid. Different people looked trustworthy. Different people looked ridiculous. Sometimes it felt as though we were not merely disagreeing about the same world. We were being shown different worlds.

And the technology surrounding us was becoming extraordinarily good at learning which version would keep each of us looking.

I kept thinking about James Madison and Federalist No. 10. Not because Madison secretly predicted the algorithmic newsfeed—please let dead philosophers rest occasionally—but because the human problem underneath it felt familiar. People organize themselves around interests, loyalties, grievances, fears, identities and passions. A pluralistic society somehow has to survive the fact that none of us sees the whole thing from exactly the same place.

Then generative AI arrived, and something about it fascinated me immediately.

Here was a machine made of language that could move between perspectives almost effortlessly.

What would a conservative say?

A progressive?

An economist?

A teacher?

A parent?

What am I missing?

What assumption am I making?

What would someone who thinks I am completely wrong notice first?

The original idea behind LBB was simple: maybe AI could help us move the view. Not tell us what to believe. Not float above democracy like some omniscient robot philosopher issuing the Correct Opinion. Just help us see that there might be more inside a question than the binary we brought with us.

Before LBB was software, it was basically a notebook for doing that.

The project went through several names because apparently I cannot leave anything named once. Life’s Banter Book. Language Beyond Boundaries. Eventually, Language Beyond Binary.

And somewhere along the way a Little Blue Butterfly wandered into the architecture.

Naturally.

But the underlying instinct stayed the same:

There is probably more here than the first frame is showing us.

AI got better.

And I used it. A lot.

The more capable the models became, the stranger and more interesting the notebook became too. A machine could help me move between disciplines with almost no friction. I could begin with a legal question and suddenly find myself reading about economics, psychology, computer science, philosophy, education, labor, history or design.

It could translate technical language. Generate competing arguments. Stage objections. Compare frameworks. Help me find terminology I did not know existed. Take some half-formed thought I could feel but could not quite articulate and hand it back with structure.

That capability still blows my mind.

Because underneath all the debates about benchmarks and model size is something very human: these systems can change what a person is capable of reaching.

And that raises a question I think is much more interesting than whether the machine itself is “intelligent” in some grand philosophical sense.

What is intelligence when human intelligence can be augmented by a machine made of language?

Where does the useful capability sit?

In me?

In the model?

In the accumulated conversation?

In the tools?

In the sources?

In the strange back-and-forth where I bring an unfinished thought, the machine returns language, I reject half of it, keep one sentence, ask a better question, and suddenly something becomes visible that neither object contained in quite the same form before?

I do not know.

That is one of the questions I want to study.

And it leads to another one.

Humans have always used language to build things that do not exist yet.

A constitution begins as language. So does a contract. A screenplay. A scientific hypothesis. A business plan. A blueprint has its own language. So does a campaign platform. A lesson plan. A court opinion. A prayer. A promise.

We represent possible futures before we build them.

That is one of the astonishing things language lets us do.

We can say: What if?

Then another mind can answer.

No, that will not work.

What about this?

You forgot the roof.

Where does the water come from?

Who pays for it?

Who gets left out?

Okay. Change the drawing.

That is planning. It is also one of the basic ways humans build together.

So now I am interested in a larger possibility:

Can we use machines that are extraordinarily good at generating language to help us imagine better futures?

And immediately behind it comes the question that now sits at the center of almost everything I am building:

How do we keep the represented future from quietly becoming THE FUTURE simply because the language describing it is convincing?

That distinction matters.

A plan can be useful without becoming prophecy.

A forecast can help us prepare without becoming fate.

A simulated worker can help us notice that workers should be consulted without becoming an actual worker.

A machine can help me reach the question an economist might ask without making me an economist.

Language gives us extraordinary reach.

Reach is not the same thing as standing everywhere we can reach.

That is where LBB changed.

The original question was:

Can AI help us think beyond binaries?

I still like that question.

But eventually I discovered the more difficult problem.

AI can generate another perspective forever.

Another argument. Another explanation. Another expert-sounding analysis. Another stakeholder. Another objection. Another rebuttal. Another possible future. Another beautifully organized explanation of why the beautifully organized explanation from six minutes ago was completely wrong.

This is both incredible and, occasionally, hilarious.

But abundance creates its own problem.

If representations become almost frictionless to produce, then judgment becomes the scarce resource.

Which answer is actually supported?

Which part is interpretation?

What has been verified?

What is merely plausible?

Who actually knows?

Who has authority?

Whose experience is missing?

What should happen next?

And—this one increasingly matters to me—when is another generated answer simply not the right next move?

Sometimes the next move is another AI answer.

Sometimes it is a source.

Sometimes it is an experiment.

Sometimes it is a teacher, doctor, engineer, lawyer, historian, worker, community, parent, government, friend or person who actually experienced the thing being discussed.

Sometimes the right answer is:

I do not know yet.

And sometimes the right move is to stop.

That became the newer version of LBB.

That brings me to languagebeyondbinary.com.

LBB is now a working prototype.

You can bring it an AI answer, claim, report, forecast, argument, passage, plan or other piece of language and ask it to help you inspect what you actually have.

What is the question?

What is the claim?

What kind of claim is it?

What appears established?

What remains inference?

What is missing?

What would change the assessment?

Who or what might actually be positioned to answer what remains?

The machine does the first analytical work.

Then LBB proposes a next move:

GENERATE · VERIFY · ROUTE · HOLD · STOP

Generate if the machine still has useful work to do. Verify if the claim needs another basis. Route if another person, source, discipline, institution or form of authority needs to enter. Hold if we do not know enough yet. Stop if continuing to generate language is not actually improving the judgment.

LBB proposes.

You decide.

That is why the line on the site is:

THE MACHINE DOES THE WORK. THE HUMAN MAKES THE CALLS.

I do not mean that humans are magically wise. We are obviously not. Have you met us?

I mean responsibility has to land somewhere.

A machine can widen the decision space. It can help us see options, expose assumptions, translate expertise, simulate possibilities and make unfamiliar questions reachable.

But eventually somebody has to decide what status the output deserves and what happens because of it.

You currently get three LBB analyses before you have to create an account.

That was intentional.

I do not want someone handing over an email address before they have any idea whether the thing is useful.

Bring it something. See what it notices. Tell it that it misunderstood you. Correct it. Try again. Give it a political claim. Give it an AI answer. Give it something painfully ordinary.

See whether it makes a distinction visible that was not obvious before.

If you want to save sessions or eventually follow a question as it leaves the AI conversation and comes back with information from outside it, then an account becomes useful.

But LBB should have to earn that relationship.

I have a daughter.

She is four.

She is hilarious.

And sometimes I look at her and realize that almost everything adults are currently arguing about when it comes to artificial intelligence is going to sound ancient to her.

She will not remember the first time a chatbot wrote a decent essay.

She will not remember people passing their phones around saying, Look what it can do.

She will not remember the collective astonishment.

Her generation is not going to experience generative AI as a shocking new invention. It will simply be part of their cognitive environment.

Machines will answer them. Teach them. Write with them. Argue with them. Explain things to them. Probably flatter them. Definitely annoy them. And increasingly, these systems will help represent the world to them.

That changes the educational question.

When a child can ask a machine to generate a persuasive explanation of almost anything, what should we teach her?

Certainly not:

TRUST AI.

And not:

NEVER TRUST AI.

Those are binaries too.

I want her generation to have more moves.

What kind of answer is this?

Fact? Inference? Forecast? Proposal? Finding? Simulation?

Who actually knows?

What evidence could answer the question?

Who has expertise?

Who has authority?

Who is affected?

Whose consent matters?

Did anybody outside the simulation actually participate?

What changed because I used the machine?

What did I learn?

What did I merely produce?

Those feel like questions worth teaching even if every model, company and interface we use today looks completely different twenty years from now.

Because the deeper educational problem is not ChatGPT. It is what happens when generating a convincing representation becomes dramatically easier than understanding what the representation has earned.

I keep coming back to that word.

Intelligence.

We talk constantly about whether machines are becoming more intelligent. Reasonable question. Important question.

But there is another one hiding right beside it:

What happens to human intelligence when humans gain access to these machines?

Does expertise become more reachable?

Does curiosity become more powerful?

Can an outsider find the right question faster?

Can a student learn something she could not reach before?

Can a lawyer cross into a technical field far enough to notice a question worth routing to someone who actually belongs there?

Can a scientist generate more hypotheses and spend more time deciding which ones deserve experiments?

Can people become better at seeing what they do not know?

Or do we become extremely good at producing artifacts that look intelligent while quietly weakening the judgment required to know whether they are?

Probably some of both.

That is exactly why I am interested.

I am not trying to build software that tells humans THE ANSWER.

I am much more interested in whether we can build tools that help humans recognize what kind of answer they have, what it still needs, and what they should do next.

And beyond even that is the question I find almost irresistible:

Can we use language to build better futures?

Not predict them.

Not generate one glorious technicolor utopia and declare victory.

Build them in the ordinary human sense.

Imagine. Argue. Compare. Question. Represent. Test. Ask who is missing. Change the plan. Bring in another mind. Discover something we did not know. Throw away the beautiful idea that failed. Try again.

Language has always helped humans do that.

Now the language answers back.

That seems worth taking very seriously.

I can continue building LBB alone with AI.

That would be easy.

It would also eventually become a pretty hilarious contradiction.

So I am looking for collaborators: educators, learning scientists, HCI researchers, AI researchers, engineers, designers, accessibility experts, journalists, historians, lawyers, people studying cognition, polarization, media literacy, labor, democracy and governance—and people from fields I do not yet know enough to name.

I need people who can tell me where I rediscovered something everyone in their field learned fifteen years ago.

People who can design better experiments.

People who can tell me “jurisdiction” is a terrible word for normal humans.

People who can show me that an interface intended to reduce anchoring actually creates more of it.

People who can build something better than I built.

And especially people willing to say:

Kat, this part does not work.

Because the work has to be allowed to lose.

Maybe students hate one of the modules. Maybe another intervention is dramatically simpler. Maybe a distinction I think matters turns out to have no measurable effect. Maybe somebody already solved part of this beautifully. Maybe the entire Board needs to disappear.

Excellent.

That is information.

There is obviously more behind why I care about language, artificial intelligence, certainty, reality, judgment, disability, and what it means to return from a story that became too certain.

That story belongs somewhere else.

My forthcoming book, A Trial of Color: Making the Case for Human Judgment in an AI-Accelerated World, comes out 11.11.26.

That is the human story.

Here, I want to build.

I want to take the odd interactive notebook I started in 2023 because I was tired of red versus blue and see whether it can become something useful for a much larger problem.

What is intelligence when a person can think with a machine?

What becomes possible when language can help us cross intellectual boundaries that once required extraordinary access?

What should we teach children who inherit machines capable of generating almost any representation they ask for?

Can AI help us reach expertise without making us believe we possess it?

When should the next move be another answer—and when should the question leave the machine?

And perhaps the biggest question of all:

Can we use this extraordinary new abundance of language to help build better futures without confusing the futures we can describe with the future that actually arrives?

I do not know.

Good.

That means there is still something worth building.

For now, go play with languagebeyondbinary.com.

You get three analyses before you need an account. Try a module. Give LBB something difficult. Correct it. Disagree with it. Tell me what it cannot see.

And if there is a piece of this you want to help build, research, teach, test, simplify—or break—I am looking for collaborators.

LBB began because I hated the feeling that two people living in the same country could open their phones and encounter two different worlds.

Three years later, AI can help us imagine more possible worlds than I could have conceived when I started.

Maybe that is the extraordinary opportunity.

Maybe it is also the harder responsibility.

The question is no longer only:

Which world is the machine showing us?

It is:

What can humans and machines imagine together?

What should we build from those possibilities?

And before the language becomes reality:

How do we decide which future has actually earned the next move?