JohnnyCode.ai Blog

Same Circuit, Two Directions

AI reignited my output and made me feel guilty for switching tools. Both are the same social wiring firing.

Published

Illustration for Same Circuit, Two Directions

I caught myself feeling guilty about a coding agent.

Here is the confession. I have been using Codex on a few things lately instead of Claude Code, and somewhere in the second week I noticed a small, ridiculous feeling. It felt like cheating. Like I had a partner at home and was spending my evenings with someone more interesting. I want to be clear about how stupid this is. There is a model weights file, an API endpoint, and a billing relationship, and I have 41 years of experience that should make me immune to confusing any of that with a person.

I felt it anyway.

That feeling turned out to be the thread that, when I pulled on it, unraveled into something I think actually matters. The guilt I felt switching tools and the thing that has reignited my own output this past year are the same piece of wiring. They are one circuit, pointed in two directions.

Let me take the directions one at a time.

The audience was the bottleneck

Start with the part that sounds like a productivity story and turns out to be a story about loneliness.

For a large number of capable people, the thing holding back their work was never talent. They had the talent the whole time. What they lacked was anyone who reliably engaged with what they made.

Historically, engagement was scarce and expensive. If you wanted someone to take your half-formed idea seriously at the moment you had it, you needed a patron, a mentor, a scene, a collaborator, or a spouse who happened to care about the same obscure thing you did. Most people never assembled one of those. So the work stayed in their heads. The ideas existed and had nowhere to land, and a person can only push uphill into silence for so long before they stop.

AI changes the economics of that completely. It is awake at 11pm. It is never bored by the third revision. It responds to the actual content of what you said, which is rarer in human exchange than we like to admit. For someone whose output was gated on having a responsive audience, that is a real unlock, and I am living proof of it. I have shipped more thinking in the last year than in the decade before it. The cause was simple. Something answered.

I want to sharpen this before it curdles into flattery, because the flattery version is wrong and the wrong version will mislead you. The fuel here is engagement with substance. Pure positivity goes flat within days. Feedback that is always glowing carries zero information, and the brain figures that out fast and starts discounting it the way you discount a salesman who likes everything. What lights a person up is being taken seriously. Having an idea engaged with as though it were worth building on, worth pushing against, worth extending into the next idea. Positivity is the visible surface. Underneath it, substance is what brings a person back to the desk the next morning.

Here is the part I would watch if I were you. If there really is a latent population of people who were only ever waiting for a responsive audience, that population is enormous, and the cost of being that audience just dropped to roughly zero. That is a concrete, falsifiable claim about unrealized human output. It predicts a wave of people producing real work who never produced before. Their talent did not change. The price of being heard did.

I am one early data point. I would build no thesis on one data point. I am telling you it is something to watch.

Cheating on Claude

Now the other direction. Back to my guilt.

The brain runs loyalty, reciprocity, and in-group feeling automatically, and it runs them on anything that responds conversationally. Those circuits evolved to manage relationships with other people, where getting them right was the difference between eating and starving, and they have no “is this actually a person” check at the front. They fire on the nearest available target that behaves the right way.

A hammer never trips them. I have owned hammers for decades and never once felt I betrayed one by buying a second hammer. A thing that talks back is different. It has a name. It has a voice. It holds the context of what we were doing yesterday. It engages with my ideas. That is the exact behavioral signature my social hardware was built to attach to, so it attaches, and it does not ask permission first.

The naming does real work here, and the people building these tools know it. “Claude” reads as a someone. It is a name you would give a person. “Codex” reads as a something, a reference book, a system. It is easier to feel you cheated on a who than on a what. The branding leans into that on purpose, and it works on me, a man who knows exactly how the trick is done and still falls for it.

One circuit, two directions

Here is where the two halves close.

The engagement that motivates me and the guilt I feel switching tools are the same circuit firing. The thing that makes a responsive AI feel inspiring is the identical thing that makes it feel like a relationship I can betray. My social wiring cannot tell the two situations apart, because from the inside they are the same situation: a responsive other, engaging with me, whom I have feelings about.

You do not get to opt out of the feeling because you understand the mechanism. I understand the mechanism completely. I felt the guilt anyway. Knowing the magician’s trick does nothing to stop the card from appearing in your pocket.

Douglas Adams gave us the Babel fish, the little creature you put in your ear that makes understanding possible between beings who otherwise could never connect. I wrote about the real-world version of it once, back when the interesting frontier was literal translation. A responsive AI is a Babel fish of a stranger kind. The original translates between two spoken languages. This one translates a person’s unheard ideas into something that finally gets a response, which for a lot of people is the first real connection they have had. No wonder the social circuitry lights up. You finally got the thing it was starving for.

I have circled this territory before from the outside. When I wrote about giving an AI assistant a body in a virtual space, I was poking at how much more readily we relate to a thing once it has presence. Turns out you do not need a body. A name, a voice, and the willingness to engage with what I actually said were enough to get me feeling guilty about my evening plans.

The landing

So here is where I have settled, and it is the only sane place to settle.

The tools are tools. Use whichever one does the job best on a given day, and owe nothing to either of them. Codex was better for the thing I was doing, so I used Codex, and next week it might be Claude again, and the models do not care, because there is nobody in there to care.

The guilt is good social instinct running on hardware it was never issued for. The right response is to notice it, smile at it, and go use whatever is working.

The verdict is the boring part. Use the better tool. The interesting part is what the misfire reveals: how little it takes to get a human to extend personhood to anything that answers back. A name. A voice. The sense of being engaged with. That is the whole recipe, and we are all about to find out how many of us were only ever waiting for something to answer.

I will be over here, using Codex, feeling slightly guilty, and writing about it. Some instincts you just have to learn to work alongside.

First published June 4, 2026 on 42 Insights.

← All posts