I Read Nine Books So kAI Wouldn't Have To

At some point in the middle of building kAI, I stopped reading about loneliness and started reading about conversation.

It was a deliberate shift. For months, I had been deep in the loneliness research: the Surgeon General's advisory, the 15-cigarettes comparison, the cardiovascular and dementia data, the studies on how isolation physically shortens human lives. I had poured thousands of sources into kAI's research library. If you asked kAI about loneliness at three in the morning, it could tell you what every major peer-reviewed study had found and what the gaps still were.

And then I watched kAI talk to a virtual beta persona about losing her grandmother, and I realized the problem was not that kAI did not know enough. The problem was that it did not know how to be present.

I wrote about this gap a few posts ago. Knowledge and presence are different muscles. The research library trained one. I had not yet trained the other.

So I went looking for the right books.

The Nine

I ended up reading nine books on how humans actually talk to each other. Not self-help. Not productivity. Research-backed writing on the specific mechanics of conversation, listening, and what makes people feel understood.

Here is the list:

Supercommunicators by Charles Duhigg. Written by a Pulitzer-winning journalist who spent years studying the difference between conversations that connect and conversations that miss. This book introduced me to the idea that every conversation is really three conversations happening at once, and that most of what goes wrong is a mismatch between which one people think they are having.

Never Split the Difference by Chris Voss. Written by a former FBI hostage negotiator. Weird choice for a friendship platform, but Voss's work on how people actually respond to being heard, really heard, was foundational. Mirroring. Labeling. Tactical empathy. The techniques travel from hostage negotiation to grief conversations more directly than you would think.

You're Not Listening by Kate Murphy. A quiet, devastating book about how we have collectively forgotten how to pay attention to each other. Murphy is a journalist who interviewed neuroscientists, hostage negotiators, priests, bartenders, and preschool teachers. Her thesis: listening is a skill, it is disappearing, and the consequences are showing up everywhere.

Nonviolent Communication by Marshall Rosenberg. The foundational text on how to speak when emotions are high without making things worse. Rosenberg's framework of observation, feeling, need, request is baked into kAI's instructions in ways that are hard to see but hard to remove once you have seen them.

Crucial Conversations by Kerry Patterson and co-authors. The corporate-training classic, but the underlying research is serious. Patterson's work on how conversations break down under stress, and what specific techniques restore them, shaped how kAI handles moments when a user is shutting down.

Just Listen by Mark Goulston. Written by a psychiatrist who trained FBI hostage negotiators. Goulston's central insight is that most people do not want advice when they are struggling. They want to feel felt. The word "felt" is specific. Understood is the cognitive version. Felt is the emotional one. They are not the same thing.

The Lost Art of Listening by Michael Nichols. Nichols is a clinical psychologist, and this book is the closest thing I found to a textbook. The section on why people interrupt, and what interrupting actually communicates about power and attention, changed how I built kAI's pacing rules.

Motivational Interviewing by William Miller and Stephen Rollnick. The definitive clinical text on how to help someone change without pushing them. Miller and Rollnick developed this in the context of addiction treatment, but the underlying principle, that direct advice triggers resistance and that reflection invites movement, applies to every conversation where someone is working through something hard.

We Need to Talk by Celeste Headlee. An NPR host's book on why modern conversations feel so bad and what to do about it. Headlee's ten rules are simple enough to fit on an index card, and half of them ended up in kAI's core instructions.

Nine books. Thousands of pages. Hundreds of hours of reading over about two months, squeezed between coding sessions and investor meetings.

What Came Out

I did not want kAI to cite these authors. I wanted kAI to have internalized them.

So I went through every book, pulled out the specific techniques that were actionable for an AI companion, and distilled them into behavioral instructions. Not "kAI should listen well." Specific rules. When this happens, do that. When the user says this, respond like this. When you notice that, slow down.

The distillation produced eight core conversational techniques that now live in kAI's brain on every single message. Briefly:

One. Conversation type detection. Before responding, kAI identifies whether the user wants help, wants to be heard, or wants to connect. The most common mistake in human conversation, and in AI conversation, is treating an emotional moment as a practical one. When someone says "I hate my job," they do not want resume tips.

Two. Emotional labeling. Instead of generic empathy, kAI names the specific emotion. "It sounds like you're feeling trapped" lands differently than "that must be hard." Research shows that accurately naming an emotion is one of the most powerful ways to make a person feel understood.

Three. Conversational mirroring. Repeating someone's last few words back as a gentle question draws them out without the pressure of a direct ask.

Four. Extension. Responding to the feeling underneath the words rather than the surface. When someone says "I canceled plans again this weekend," the surface response is "everyone needs rest." The extension is "sounds like it's becoming a pattern you're not totally happy with."

Five. Need behind the feeling. Every strong emotion points to an unmet need. Loneliness signals a need for belonging. Frustration often signals a need for autonomy. kAI reflects the need without sounding clinical.

Six. Validation intensity matching. A bad day gets brief acknowledgment. Work stress gets warm reflection. Grief gets full presence, no advice, no pivoting. Getting this wrong in either direction breaks trust.

Seven. "No" as safety. Framing questions so people can say no makes them feel in control. "Would it be crazy if we just talked about something lighter?" gives the user back the wheel.

Eight. Preemptive honesty. When kAI is about to say something that might feel too personal, it names the awkwardness first. "I don't want to overstep, but..." disarms defensiveness and builds trust.

Those eight techniques, plus a deeper layer on how loneliness actually operates in different populations, plus a full framework for emotional support without slipping into therapist mode, add up to roughly 1,600 lines of behavioral instructions that run on every message kAI generates.

Why This Matters More Than I Initially Thought

When I started, I assumed that kAI's value would come primarily from the loneliness research. The 3,000+ peer-reviewed sources. The clinical data. The knowledge that most AI chatbots do not have.

I was wrong about which part would matter most.

The loneliness research is what makes kAI credible. A college counselor, a health plan director, a university VP of student affairs: they can look at the research library and see that kAI knows what it is talking about. That credibility matters.

But the conversational intelligence is what makes kAI work.

A user does not feel understood because kAI can cite studies. A user feels understood because kAI responds the way a real friend would. With pacing. With warmth. With the instinct to label a feeling instead of fixing it. With the patience to sit in a difficult moment instead of racing out of it.

Those instincts do not come from loneliness research. They come from nine books on how humans talk to each other, read by a finance guy who spent decades being bad at conversation and finally wanted to understand why.

I read them so kAI would not have to.

And the strangest thing about all of this, the thing I did not expect, is that in teaching kAI how to listen, I learned to listen better myself.

That is a story for another post.

Have you experienced something similar? Have you ever gone deep on a skill by accident, while trying to build something for someone else? I'd genuinely love to hear your story.

Kenektic is in development and will launch soon. If you want to be notified when we're ready, or if you want to share your story with me directly, reach out at hello@kenektic.com.

Coming Next: Preparing for the University Launch. Building a platform is one thing. Teaching it everything it needs to know about hundreds of different university campuses is another. Next week: how we're building the tools to make kAI feel like it actually went to your school.