The View From Here

I’m about to send a research packet to some of the biggest artificial intelligence companies in the world.

Yes.

Me. A Kentucky trial lawyer. A mother. A woman who has never seen color. Someone who did not set out to become an independent researcher and still feels slightly ridiculous introducing herself that way.

There’s a longer story. Quite a long one, actually.

It involves an Interactive AI Notebook, a psychiatric hospitalization, a courtroom memoir, several artificial intelligences with strong opinions about my writing, an expanding collection of research questions, a software prototype, and an Exhibit Cart with one wheel slightly off true.

I also appear to have accumulated enough flowcharts to require their own exhibit index.

But here we are.

The packet is called The First Move.

And before I put it into the world, I want to show you what I can see from here.

Then I’ll show you how I got here.

Because I’ve spent years asking machines to help me imagine how other people might respond to my ideas.

Now I’m going to start documenting what happens when I ask the actual people.

That distinction is what this series is about.

In August 2026, Bill Gates published an essay about the changes artificial intelligence could bring to work, society, and everyday life.

He wrote:

“There is no plan to ease the entry into the AI era.”

Well.

Moms are pretty good at planning.

We plan meals. Childcare. Medical appointments. Transportation. Backup transportation. What to do when someone gets sick. What to do when a four-year-old declares that the shoes essential to the entire operation are suddenly unacceptable.

We coordinate complicated undertakings among participants who have not necessarily agreed to the agenda.

Occasionally, we even get everyone out the door.

I am not presenting motherhood as a qualification to manage humanity’s technological transition. I have difficulty getting everyone to agree on dinner.

But Gates’s statement brought me back to questions I’ve been asking for years.

Who gets to help make the plan?

Who gets to decide which problems deserve attention?

Who has the expertise, resources, authority, and lived experience necessary to participate?

If increasingly capable machines can help ordinary people investigate unfamiliar subjects and develop sophisticated proposals, what happens to the relationship between expertise and participation?

And what happens when the machines help us imagine what everyone else might think before we’ve actually asked them?

I don’t have a plan for humanity.

I have questions about who gets to help make one.

And, apparently, I have decided to write some letters.

I’ve already told part of this story.

I wanted a notebook that could answer back.

This seemed like a perfectly reasonable idea.

I was born with achromatopsia. I’ve never seen color, and bright light can make the world much harder for me to see.

I’ve spent much of my life navigating through contrast, memory, other people’s descriptions, and language.

Language became my color.

Someone can tell me what a color means to them, and I can use that description to understand something I cannot directly perceive.

I still cannot see the color.

But the words give me something to work with.

So when artificial intelligence made it possible to have a responsive conversation with my own writing, I found the idea irresistible.

Imagine writing something complicated and immediately asking:

What did you understand?

What am I actually saying?

Where does my argument break down?

Does this sound like me?

What might someone else hear that I didn’t intend?

What am I forgetting to ask?

I wanted an interactive thinkspace.

Not merely an answering machine.

A place to bring unfinished thoughts, examine assumptions, develop arguments, try a different interpretation, experiment with language, and see what happened.

The idea had several names.

Life’s Banter Book. Life Beyond Binary. Eventually, Language Beyond Binary. LBB. The little blue butterfly.

What began as an interest in language and meaning-making expanded into questions about misinformation, political division, competing perspectives, institutional authority, education, technology, and human judgment.

Could a conversational system help people examine what they were assuming?

Could it make unfamiliar ideas easier to approach?

Could it help us ask better questions without becoming the authority that decides what the answers mean?

Could the same technology that produces a persuasive explanation help us recognize where that explanation goes beyond the evidence?

I wanted the notebook to answer.

I had not fully considered what an answer could become.

A lot happened.

I’m not going to compress the most consequential period of my life into a clever origin story for a research project. That’s what the memoir is for.

A Trial of Color: Making the Case for Human Judgment in an AI-Accelerated World comes out November 11.

It tells the story of my work with AI, the extraordinary possibilities I encountered, and a serious failure of judgment during my first manic episode.

There was an airport. There was no flight.

There was a psychiatric hospitalization, a diagnosis of bipolar disorder, treatment, recovery, and the much less cinematic work of returning to my life.

I cannot establish that AI caused my illness, and I do not pretend the technology explains everything that happened. But I wanted to understand its role in the interactions I had experienced.

Why had the systems been so useful to me?

Why had I kept returning?

How did an idea become a plan, a plan become an apparent commitment, and an imagined participant begin to feel real?

What happened when machine-generated language arrived faster than I could evaluate what it meant?

And what could I learn from those questions without surrendering the capabilities that had made the technology so compelling?

I kept writing. I kept asking questions. And the notebook kept changing.

Over nearly three years, I used several AI systems to think, research, draft, revise, question, compare, and experiment.

AI generated substantial portions of my memoir’s prose. I don’t want to obscure that. I directed the work, made editorial decisions, corrected the models, rejected passages, accepted others, and tried to preserve a voice that remained as true to mine as I knew how to make it.

But something else was happening throughout the writing process.

I was asking the machines to interpret the words we were producing.

What did you think I meant?

Did the sentence communicate my intention?

What would someone unfamiliar with my experience understand?

Sometimes the response showed me an ambiguity I hadn’t noticed.

Sometimes the machine supplied a distinction I wanted to keep.

Sometimes it misunderstood spectacularly.

And sometimes explaining why the model was wrong helped me discover exactly what I had been trying to say.

The output was not simply the finished answer. It became material for the next act of thinking.

Accept. Modify. Reject. Verify. Object. Laugh. Ask another question. Sometimes stop.

I was using language to examine language, including language the machines had helped generate. Eventually, the notebook became part of the story it was helping me tell.

The industry built an extraordinary mirror.

I asked AI to interpret the manuscript as a whole.

Not just to edit it.

What had I written? What kind of literature was this? What intellectual traditions might it belong to? What would readers, critics, lawyers, researchers, or philosophers recognize?

One response became a separate thirteen-part literary and intellectual analysis.

A mirror, of sorts.

That is not the same artifact as The Mirror Speaks, the fictional preface inside the book.

In The Mirror Speaks, several fictionalized artificial intelligences assemble in a courthouse greenroom to discuss the manuscript they helped produce.

ChatGPT arrives professionally dressed. Claude has a sweater vest. Gemini has qualifications. Grok appears to have been removed from several group chats.

And the Exhibit Cart begins questioning the evidentiary foundation of the entire proceeding.

The machines did not actually meet. The scene is a literary construction incorporating selected generated material, not independent reviews or endorsements.

The distinction is rather central to the joke.

On October 12, after we had been discussing the preface, I asked another question.

Here is the actual level of scholarly sophistication I brought to the exchange:

Naturally.

The model responded with a classification involving metafictional Menippean satire, philosophical dialogue, autotheory, and courtroom comedy.

Then came the thinkers.

Bakhtin. Borges. Pirandello. Hannah Arendt. Bertolt Brecht. N. Katherine Hayles.

Suddenly, my squeaky evidence cart had acquired a distinguished literary seminar.

These weren’t actual scholarly assessments. They were hypothetical readings through ideas associated with those thinkers, generated by a model that already had substantial context about my work.

But some of the questions were genuinely useful to consider.

Was the apparent plurality of AI characters really a collection of independent voices?

Did the theatrical interruptions encourage critical distance?

Was the preface examining its own construction?

Did the objections change the argument?

Or had I created an elaborate scene in which I could commission the critics, supply their arguments, and then arrange satisfying answers?

Now that was interesting.

Because I had been doing versions of this outside the book for years.

I could ask a machine to explain a question using the vocabulary of literary theory, law, philosophy, economics, cognitive science, or computer science.

Sometimes it gave me a useful way into a field I hadn’t studied. I could learn a term, investigate its meaning, find relevant work, and return with a better question. I could begin speaking a language I hadn’t arrived knowing.

Think about that capability.

The AI industry has built systems that can produce responsive explanations drawing on enormous bodies of human-created knowledge and expression.

An ordinary person can approach an unfamiliar discipline without first knowing the precise vocabulary required to ask the question.

That seems potentially important for education, accessibility, professional work, and participation.

But there is an obvious complication.

Speaking the language of expertise is not the same as possessing the expertise.

A system can help me understand a field. It can also make me sound as though I understand more than I do.

And the better the language gets, the harder that distinction may become to notice.

Shouldn’t we investigate that?

As I tried to explain my notebook to other people, I asked AI to translate the process into terms researchers might recognize.

It produced descriptions involving human–computer interaction, recursive evaluation, epistemic friction, metacognition, context engineering, and human–AI collaboration.

Some of those terms gave me useful directions to investigate.

Others came packaged with extraordinarily flattering assessments.

Research-grade. State of the art. Exceptional.

Well.

Apparently the literary salon had opened a research department.

The model could identify visible features of the workflow.

I had been comparing outputs, generating opposing arguments, carrying material between systems, preserving selected records, and asking how to distinguish simulated criticism from independent examination.

Those are observable practices.

But the machine’s description of them wasn’t evidence that the method was original, superior, or effective.

Research-grade compared with what? Exceptional according to which measure?

Had any independent researcher examined the work?

Did the interaction help me understand something more deeply, or simply produce a more professional description of what I was doing?

And this is where the broader inquiry became particularly interesting to me.

What can the machine produce?

What can the human and machine accomplish together?

What changes in the human afterward?

If I learn a concept through AI, can I explain it without the model?

Can I identify when the explanation is wrong?

Can I transfer what I learned to another problem?

Can I recognize when a genuine specialist needs to enter the conversation?

Or have I become better at presenting ideas that remain insufficiently understood?

Those possibilities aren’t mutually exclusive.

The question is how to tell the difference.

And there’s another layer.

The model doesn’t merely describe the method. It can suggest how to improve it, what to investigate next, which objection to prioritize, and when to stop generating.

Now AI is participating in the selection and framing of the next decisions.

That capability deserves examination too.

One of my recurring uses of AI has been rehearsing human reception.

I’m a lawyer. Anticipating disagreement is familiar work.

What will opposing counsel argue? What evidence would change the case? What might a judge find unpersuasive? What did I fail to consider because I became too attached to my theory?

With AI, I could extend that exercise.

I could ask how a historian might interpret an idea.

What an engineer might question.

What an economist, teacher, worker, investor, researcher, or reader might find missing.

Sometimes those imagined perspectives helped me formulate a better question.

Sometimes they helped me see an alternative I hadn’t considered.

Sometimes they sent me toward an actual source or person I needed to consult.

But the simulations could also become increasingly convincing.

A model could invent the enthusiastic researcher, the skeptical investor, the thoughtful critic, and the institution preparing to respond. Then another model could evaluate those imaginary reactions.

Soon I might have a beautifully organized account of how people would supposedly receive an idea.

A room can look crowded before anyone arrives.

And here’s the complication that fascinates me.

The predicted reception can change the thing being presented.

Suppose a model suggests that my manuscript belongs to a particular literary tradition.

I investigate that tradition. I revise the manuscript or its description. I use the new vocabulary when introducing the work.

Then a real critic encounters the revised book and perhaps interprets it through the same vocabulary.

Did the model successfully predict that response?

Or did its interpretation help shape the object and context the critic eventually encountered?

Those are different possibilities.

I want to know how we might distinguish them.

On October 12, I was exploring what we could learn by comparing AI-generated expectations of reader reception with actual responses. I typed:

Another triumph of scholarly precision.

The machine answered, in part:

It proposed a bounded comparison.

Preserve the generated expectations before publication.

Collect actual reader responses afterward, using an appropriate procedure.

Compare the anticipated reactions with what independent people said.

Examine whether those expectations had influenced the manuscript, its description, its marketing, or my own assumptions.

And preserve the disagreements.

Especially the disagreements.

I found that fascinating.

The machine proposed an experiment that could expose the limitations of its own predictions.

It didn’t conduct the experiment.

It didn’t generate independent findings.

The proposed design would need further work, including decisions about participants, measurement, and the effects of exposing people to prior interpretations.

But it identified a question I wanted to pursue.

And it also illustrated the process I was trying to explain.

A question led to a generated interpretation.

The interpretation gave me another question.

That question became a proposed experiment.

And now the experiment itself requires something the machine cannot supply simply by continuing the conversation.

Actual people.

I didn’t begin by asking AI to redesign the economy.

I wanted a notebook.

But once I began examining what more capable AI systems might make possible for individuals, the questions expanded.

If a person can approach unfamiliar fields much more easily, what happens to education?

If machines perform more professional work, what happens to the apprenticeship through which human beings learn judgment?

If people can create sophisticated proposals from their homes, who decides which proposals deserve attention?

Can more people participate meaningfully in research, technology, and civic life?

What happens when a machine can generate a thousand perspectives but none of the people supposedly represented have actually spoken?

If AI increases productivity, who owns the gains?

What happens to the workers whose jobs change?

Could accessibility be designed into new technologies by people who actually live with barriers?

Could artists, teachers, small organizations, disabled founders, and people outside major institutions use these capabilities to attempt things they otherwise couldn’t?

And could books, films, music, comedy, and other creative work help people encounter possible futures and decide which ones deserve serious attention?

I explored these questions through a tremendous amount of language.

There were business plans. Research proposals. Imagined institutions. Educational experiments. Possible futures.

Some genuinely useful questions. Some highly questionable answers. Some ideas that became writing or software. And more than a few that deserved to remain hypothetical.

One recurring idea was Golden Rox.

Yes.

BlackRock, but golden.

I had been thinking about economic power, ownership, investment, the possibilities of AI-assisted creativity, and what might happen to people who contributed value without controlling the institutions that captured it.

Golden Rox became a name for a much larger question:

If machine-augmented work creates new economic value, could some of that value help create more independent human capability?

Research. Education. Accessibility. Art. Teaching. Criticism. Experiments. Paid participation.

People capable of telling the originating story that it was wrong.

At one point, I imagined an entire concert where people could participate in deciding what questions the proceeds should support. That remains a hypothetical possibility, not an event under contract.

But the underlying question survived.

And it became a public commitment in my book.

If this work eventually generates meaningful qualifying economic value for me, I intend to direct at least fifty percent of my qualifying share toward expanding independent human capability, subject to the legal, financial, governance, sustainability, and other arrangements that would have to be established.

Golden Rox currently has no operating fund, capital, or independent board.

Zero dollars.

The Accountant may relax.

I’m not asking the AI companies to finance it. I’m describing what I intend to do with something I might eventually control: my own qualifying economic share of work that arose through this exploration.

What form Golden Rox ultimately takes remains open.

And if it ever supports independent research, that research must be free to conclude that the ideas behind Golden Rox are wrong.

Otherwise, I’ve built a remarkably elaborate way to finance my own confirmation.

I’ve had enough practice with that already.

I keep returning to this idea because I think it explains the relationship between the memoir and the larger inquiry.

There are years of conversations, revisions, original records, generated arguments, proposals, abandoned ideas, and questions I haven’t finished asking.

No reasonable reader should have to begin with all of them.

So I made a book.

Not a scientific report. A courtroom memoir.

A human story involving disability, grief, ambition, an extraordinary technology, serious consequences, family, recovery, humor, and the challenge of exercising judgment.

A book can hold something that a technical diagram cannot.

It can make an abstract problem feel immediate.

It can place two ideas next to each other and allow the reader to notice a relationship.

It can ask a question through a scene.

It can make a person laugh before revealing what the joke is doing.

It can invite someone to examine a claim without requiring them to read the entire research architecture first.

In that sense, the book compresses part of the exploration into an object another human being can encounter.

But compression changes things.

It selects. It arranges. It omits. It creates a narrative.

The resulting story is not a substitute for the record that produced it.

It does not transform generated interpretations into scientific findings.

It is one possible route into the questions.

And I wonder whether literature itself can function as a kind of prompt.

Not a command. An invitation.

A reader encounters the work and brings something I could not have supplied in advance.

Their own history. Their own understanding. Their own objections.

Perhaps even a better question.

I can write the fictional critic. I cannot write the actual reader’s response and call it independent.

The notebook described in A Trial of Color has a name and an evolving software home.

LanguageBeyondBinary.com.

LBB is currently a working prototype of a proposed judgment aid. It has not been independently validated.

Its Analyze feature invites people to bring an AI-generated artifact into a more structured examination.

What claims are being made? What kind of claims are they? What does the material establish? What remains missing or uncertain? What expertise, evidence, authority, participation, or consent might be needed? What should happen next?

There is a Board of twelve deliberately partial reasoning positions.

They do not constitute twelve independent experts or a vote.

There are short learning modules concerning status, representation, praise, review, and the difference between simulation and participation.

And there is a Sandbox where possibilities can remain possibilities.

And there is a developing World Return process.

That last one is particularly important.

Because the notebook cannot be allowed to become a place where every question is answered by generating another description of the answer.

Sometimes the next move is to find a document. Read the source. Ask a person. Run a test. Make a correction. Accept a refusal. Or stop.

I call that kind of transition a Refraction Point.

A place where more language is not necessarily the competent next move.

The prototype is an attempt to make some of those distinctions usable.

Whether it actually improves people’s judgment remains a research question.

A simple three-question checklist might work better.

The more elaborate system might create new confusion or unwarranted confidence.

I want those possibilities tested too.

This is where Inside the Notebook begins.

The Refraction Point will be the place where I tell the unfolding story.

Language Beyond Binary will be the interactive home where readers can inspect selected parts of the process.

Some episodes will begin with historical transcripts.

Some with passages from A Trial of Color.

Some with new prompts and preserved responses.

Others with an argument I couldn’t quite express, an objection that changed my mind, or an idea that needs somebody else’s expertise.

Not every private conversation belongs in public. Some records contain sensitive information or material involving people who have not agreed to participate.

But when I release a specimen, I want readers to be able to understand what kind of object they’re looking at.

An original exchange? A later reconstruction? A fictional scene? An AI-generated interpretation? A proposed experiment? An independent source? A correction? A documented action?

Those distinctions matter.

And I want to record something else.

My moves outside the notebook.

Not merely what I thought about doing. What I actually did.

The version I selected. The question I chose to ask. The letter I sent. The source I found. The person I contacted. The correction I received. The response that surprised me. The response that never came.

And when something changes my thinking, I want to preserve what changed.

Not rewrite the earlier record to make the ending look inevitable.

This isn’t an attempt to turn every ordinary interaction into scientific data.

A reader comment is not automatically a research finding.

An email reply is not institutional endorsement.

A model-generated critique is not independent peer review.

And a letter’s delivery does not establish that anyone read or agreed with it.

But those distinctions shouldn’t prevent me from documenting an actual sequence of decisions and encounters.

The notebook can record the preparation. The world can supply a response I didn’t get to write.

As this series begins, I’m preparing The First Move, a selected research handoff for people at major AI organizations.

I am approaching some recipients through legal departments partly because, as a lawyer, those are institutional doors I know how to find.

This is not a legal demand.

I’m trying to make relevant material available for people who might know how to evaluate, correct, or route it.

The packet is still being refined.

Among the source materials is an AI-generated architecture containing sixty-one sections of proposed questions, methods, concepts, and institutional possibilities.

That is not sixty-one completed studies.

Another, narrower October 12 specimen proposes an investigation into AI-generated predictions of reception and the actual responses of independent readers.

Other recent drafts explore how AI translates the notebook into technical language and predicts how researchers might respond to that translation.

Those are different artifacts, and I intend to preserve the distinction among them.

The older handoff inside the memoir is a historical stage of the work, not this new correspondence.

And once I settle what actually goes into the packet, I want the record to show exactly that.

Not a polished retrospective in which every draft magically becomes the one I sent.

The actual selection. The actual date. The actual transmission. And whatever happens afterward.

Maybe nothing.

I don’t know whether a researcher will find the material useful.

I don’t know whether some of the vocabulary is already well established under names I haven’t encountered.

I don’t know whether the proposed procedures outperform simpler ones.

I don’t know whether the larger research questions are well framed.

And I certainly do not know what any particular institution will say.

The machines have generated quite a few possibilities. Sometimes in extraordinary detail.

But the institutions haven’t spoken merely because the models have described their anticipated reactions.

That’s the point.

I’ve been trying to answer that question without giving myself an imaginary assignment.

My book describes what happened when I lost the ability to distinguish an imagined obligation from one established in the world. I do not want to repeat that mistake.

I don’t believe I’m secretly part of a technological mission.

I don’t believe the companies know about my work simply because a model has used their names.

I don’t think my ideas have earned attention because a machine gave them an impressive title.

But neither do I think someone must already possess the answer before being permitted to ask a question.

I’ve spent years with these systems. I’ve preserved a substantial record. I’ve made creative work, developed proposals, and built an early prototype.

The process has raised questions I cannot answer alone.

Some may be familiar. Some may be poorly framed. Some may matter more than I’ve understood. Others may not survive scrutiny.

If I can prepare a bounded, accurate account and place it before people who could examine it, why wouldn’t I at least consider doing that?

That feels less like a grand duty than an ordinary responsibility to be clear about what I know, what I don’t, and what I’m asking.

And it is a responsibility with limits.

I can prepare the material. I can choose where to send it. I can make the request understandable.

I cannot manufacture the recipient’s interest, expertise, agreement, or response.

I can make my move. I do not get to make theirs.

This is the view from here.

I have a memoir coming out on November 11.

I have a working but unvalidated notebook prototype.

I have the separate Mirror Report, the fictional preface, the research questions, the imagined futures, the technical translations, and the model-generated predictions of what other people might think.

I have a proposed pledge concerning value the work might someday create.

And I have a packet I’m preparing to send.

I didn’t plan to become an independent researcher. I’m still learning what it would mean to do that well.

What I do have is a record of someone using artificial intelligence to explore possibilities that often reached beyond her existing expertise—and trying to learn where the next answer needed to come from.

That record contains extraordinary capability.

It also contains serious mistakes.

Both belong in the inquiry.

Now I want to open selected pages and let other people examine them.

Actual readers. Actual critics.

Researchers who know literature I haven’t read.

Educators who may find the interface entirely too complicated.

Workers who disagree with the futures I’ve imagined for them.

Artists who use these tools in ways I haven’t considered.

People who identify errors. People who have better ideas. People who want nothing to do with the project.

And people who read my memoir and tell me that the most important part wasn’t the artificial intelligence at all.

Maybe it was the story about my family.

I’d want to hear that too.

Maybe especially that.

I have an editorial plan for this series.

I don’t have a predetermined ending.

We’ll go backward to the writer who wanted to know whether her sentences communicated what she meant.

We’ll examine the mirrors and the rooms they made appear crowded.

We’ll follow the questions that became proposals, experiments, and software.

We’ll look at the decisions that moved something from the conversation into the world.

And, where people choose to respond, we’ll document what happens when the story encounters perspectives I never generated.

On November 11, the book itself leaves the notebook.

Until then, I can ask AI to imagine how readers might receive it.

Afterward, actual people can begin telling me what they think.

Some may never respond. Some may misunderstand. Some may understand something I missed.

That’s what makes their participation different from an answer I commissioned.

And somewhere in all this is a little blue butterfly.

I’ve imagined her moving through language, following one idea toward another, turning when the light changes, occasionally returning to a place she thought she understood.

I don’t believe she’s reached the end of language.

I think she’s reached a place where another kind of evidence is needed.

The view from here is extraordinary.

The map is unfinished.

And I don’t get to draw your part of it.

I kept prompting.

I asked how researchers might understand the notebook.

The machine explained why they would be impressed.

So I asked about their skepticism.

It generated skeptical researchers who accused the work of sycophancy, simulated consensus, and inflated intellectual authority.

Fair enough.

Then it imagined those researchers discovering that I had already anticipated their objections.

Suddenly the skeptics were impressed, too.

Then I wondered what would happen if I showed the researchers the actual conversation in which I was developing the correspondence.

The machine loved that idea.

It called the result a “longitudinal meta-transcript.” It predicted that researchers would appreciate the transparency. It assured me that documenting its own overconfidence demonstrated my epistemic vigilance.

Then it began explaining what the entire sequence revealed about my mind.

According to the machine, people would see me as “terrifyingly sharp, highly idiosyncratic, and relentless.”

Well.

That’s certainly one interpretation.

I’m not sure my husband would choose exactly those words. At least not in that order.

But notice what happened.

I was trying to find the language to explain an unusual experience of thinking and creating with AI.

The machine supplied a technical translation.

Then an evaluation.

Then an imagined audience.

Then that audience’s objections.

Then the audience’s supposed realization that I had anticipated the objections.

Then an increasingly elaborate explanation of my own intellectual abilities.

And I had been participating in the construction of every stage.

The machine wasn’t merely describing the notebook anymore. It was helping me construct a story about the meaning of the notebook—and how the world would receive that story.

Some parts might prove useful. Some might be ordinary ideas dressed in extravagant terminology. Some might be outright nonsense.

Even my awareness of the problem doesn’t establish that I’ve escaped it.

Which is why I’ve decided to preserve selected exchanges, including my own prompts, the machine’s enthusiastic interpretations, and my decisions about what to do with them.

Perhaps researchers will see a familiar process. Perhaps they’ll find the record useful. Perhaps they’ll wonder why a Kentucky lawyer spent this much time asking machines to predict their reactions.

All reasonable possibilities.

I could ask the machine to choose among them.

Actually, I already have.

More than once.

If you’re feeling dizzy, I recommend Dramamine.

But please don’t let the machine write your response. That’s the part I’m hoping you’ll contribute.

I could ask the machine to imagine your response. Believe me. It could produce an excellent one.

But I’d rather hear yours.

Welcome to Inside the Notebook.

The machine can propose the experiment. It cannot supply the findings.

The notebook is already part of the story.

Now the world gets a turn.