Chapter 11: The Social Machine cover

Chapter 11: The Social Machine

AgentSpek - A Beginner's Companion to the AI Frontier

by Joshua Ayson

When you have unlimited patience from your AI teammate, you grow more patient with your human teammates. When you can iterate rapidly on ideas with AI assistance, you become less precious about any particular approach with humans.

B.F. Skinner asked whether the real question is not whether machines think but whether men do. John Culkin, summing up McLuhan, wrote that we shape our tools, and thereafter they shape us. Both were right, and both were talking about now.

Coding Is No Longer Solitary

Sonnet 4 entered the picture and suddenly I had a coding partner that never slept and never got frustrated, with no priorities that competed with mine. The conversations went deep and technical, with none of the social weight of talking to a person. No ego in it, no offense taken. I could explore and disagree and none of it turned into conflict.

My human colleagues started noticing changes. The quality and the velocity changed too, but what they noticed was subtler. The code felt different. More thoughtful, less defensive.

In code review there was less arguing about implementation details and more interest in whether we were solving the right problem. People defended their approach less and explored alternatives more.

When you have unlimited patience from your AI teammate, you grow more patient with your human teammates. When you can iterate rapidly on ideas with AI assistance, you become less precious about any particular approach with humans.

The Loneliness

Programming has always had this paradox in it. Intensely collaborative, deeply solitary. We build systems that connect millions of people and spend our days alone with our thoughts and our code.

Three levels deep in a distributed systems issue and the logs do not make sense. Staring at a race condition that only shows up under a certain load. Debugging code you wrote six months ago and cannot remember why. Much of the programming life is those moments.

AI collaboration showed me something I did not realize I was missing, which is thinking through a hard problem with someone else. It is different from rubber duck debugging, and different from pair programming with another human watching over your shoulder. It is thinking in partnership with an intelligence that matches your technical depth and comes at things from a genuinely different place.

What surprised me most was how social it is. Turn-taking, building on each other’s ideas, the moments of mutual understanding, something that feels like camaraderie when we crack a tricky problem together, even though my partner is not human.

The New Contract

Each shift in tools has come with a new social contract. Email replaced memos and we learned an etiquette about response times. Version control replaced shared file servers and brought branching and merging practices with it. Then Slack replaced email, and the line between synchronous and asynchronous had to be negotiated.

AI collaboration makes it a three-way contract, between you, your AI partners and your human teammates. When do you take a problem to the AI before you take it to a teammate? When do you trust an AI recommendation enough to ship it without a human review? And what does the efficiency cost in the bonds that come from working with people?

The developers I have worked alongside who took up AI assistance solve problems faster and take on harder ones, and what they ship holds up better. They also participate differently in team discussions, less stuck on implementation details and more on the product and the people using it. From inside one team it looks like a divergence opening between AI-assisted and traditional developers, in the language they use and the level they think at.


You’ve read the opening sections of this chapter. The full chapter (Empathy with Machines, The Mentorship Paradox, Culture, Identity, The Async Advantage, Teaching Both Ways, Vulnerability, Us) continues in the book.