The Longer Way Around

Satoshi had a small fear he hadn’t told anyone.

Not about being laid off, not about money — those were real but manageable. This one was harder to explain.

It had started a few months ago, when he’d noticed that he couldn’t quite hold a technical problem in his head the way he used to. The edges kept softening. He’d reach for an argument and find the AI had already made it. He’d start to form an opinion and realize he was just paraphrasing something he’d read back from a prompt.

He wasn’t sure if the tool was doing it, or if it was just what happened when you stopped being in the work every day. But the fear was there, quiet and specific: What if I’m getting dumb?

The consulting engagement came from a former colleague — a workflow audit, three days of work. Small enough to be manageable. Technical enough to matter.

Satoshi opened the documentation, opened the AI tool almost in the same motion, and typed: Here’s a workflow. What’s wrong with it?

Twelve seconds. Four findings, two low severity, one medium, one worth flagging. Clean, organized, correct.

He read it twice.

Then he typed: I don’t think your answer is right.

It wasn’t a considered position. He didn’t actually know yet whether it was right or not. He just didn’t want to accept it so easily.

Which part? the AI returned.

The retry mechanism. You called it low severity. I don’t think it is.

The AI explained its reasoning. Satoshi read the explanation, thought about it, and pushed back again — not because he’d found a flaw, but because he wanted to understand it well enough to either agree or disagree on his own terms.

What’s the failure scenario you’re not mentioning?

What would this look like at 3am with no one watching?

If the retry loop triggered fifty times overnight, who would know?

The AI worked through each one. Some of its answers held. One didn’t — a gap in the logging it had underweighted. Satoshi wrote it down in his notebook, not the AI’s words, his own: No visibility when it fails quietly. That’s the real problem.

He looked at that sentence for a moment. He’d figured that out. Not retrieved it — figured it out, by refusing to stop at the first answer.

He kept going for two hours. Not accepting, not rejecting — just pushing. Every time the AI gave him something, he asked what it was missing, what it assumed, where it would break. His notebook filled up. Some of it contradicted itself. He crossed things out.

By the end, he was tired in a way he hadn’t been in months. Good tired.

He submitted the audit the next morning.

The reply came from a name he didn’t expect: David Lim. Not the colleague who’d brought him in — his old manager. Apparently the engagement had been routed up. David ran the infrastructure division now.

This is sharp work, Satoshi. Especially the retry issue — the last team missed that completely. Can we find time to talk this week?

They talked on Thursday. David was warm, direct. There was a role, senior level, the kind of thing Satoshi would have said yes to without thinking two years ago. Good problem space. Real ownership. The team was solid.

Satoshi listened to all of it.

“I’m going to pass,” he said. “But thank you. I mean that.”

David paused. “Can I ask why?”

Satoshi thought about it. He could have said he was exploring other directions, or that the timing wasn’t right. Both would have been partially true.

“I think I needed to prove something to myself first,” he said. “And I think I just did. I don’t want to stop now.”

David was quiet for a moment. “That’s a good answer,” he said. “Keep in touch.”

After the call, Satoshi sat at the kitchen table without opening his laptop. The notebook was there, covered in crossed-out lines and circled fragments. He didn’t read it. He just left it open.

Outside, someone was walking a dog, slowly, letting it stop at everything. No hurry. Satoshi watched until they turned the corner.

Then he opened a new page in the notebook and wrote one question at the top.

He didn’t answer it yet. He wanted to think about it first.

Think of the last time you accepted an answer — from a tool, a search, a colleague — without really testing it.

What would have happened if you’d pushed back, just once?

This week, try disagreeing with the first answer you get on something that matters. Not to be contrarian — just to see what’s underneath.

Tomorrow’s Sidekick is a quiet story series about staying human as the world speeds up.
New episodes every week. If this resonated, feel free to forward it to someone who’s trying to stay human in the age of AI.