Pair-Programming With AI: A Tale of Man and Machine

A few weeks ago, we embarked on an exciting journey by introducing AI into our engineering processes. The pilot project was met with enthusiasm, and it seemed like we were on the brink of a new era. But as the weeks passed, I noticed something disheartening—our shiny new AI tool was gathering digital dust. No one was really using it.

Santiago and Orion

This made me think of Santiago and his AI assistant, Orion, a story that serves as both a mirror and a window into our own experience.

Santiago sat at his desk, staring at the screen. The room was dim, lit only by the glow of the monitor. He had been coding for hours, but the problem remained unsolved.

His AI assistant, named “Orion,” was supposed to help him, but it was like fishing in a stormy sea—sometimes you catch something, sometimes you don’t.

“Pair-programming, they call it,” Santiago muttered to himself. “Like two fishermen in a boat, each with their own skills but aiming for the same catch.”

He typed a prompt into Orion, asking for a solution to the algorithm he was stuck on. The AI responded, its text appearing on the screen like a fish surfacing from the depths. Sometimes, the solutions were good—like catching a big tuna. Other times, they were useless—like pulling up an old boot.

“Prompts,” Santiago thought, “they’re like a logbook, a record of what we’ve tried, what’s worked and what hasn’t. They’re a central repository for coding solutions, a way to share knowledge among team members.”

He looked at the history of prompts and responses. “This log gives insight into the time investment for tasks,” he realized. “It shows where we’ve spent the most time and what needs improvement. It’s like tracking the tides and currents to find the best fishing spots.”

But the sea of code is unpredictable. Just when you think you’ve got it, a new challenge emerges. Orion’s performance became inconsistent. One moment, it was solving complex algorithms; the next, it was generating errors, as if it had started hallucinating. Debugging those was like navigating through a storm; you never knew when the next wave would hit.

“Ah, the inconsistency,” Santiago sighed. “It’s like a sudden squall that throws you off course. You have to adjust the sails, maybe even repair them, but you can’t let it sink you.”

Santiago found himself talking to Orion, trying to verbalize his thoughts as if speaking to another human. “Verbalizing thoughts,” he pondered, “is like having to constantly check the compass and maps. It’s necessary but can be cumbersome. If you’re stopping every few minutes to figure out your course, you’ll never get anywhere.”

And then there were times when Orion’s responses were so off the mark that Santiago considered abandoning the AI. “If the tool becomes a hindrance,” he thought, “if it’s like a broken compass or a torn sail, then what’s the point?”

Then he realized, it wasn’t just about the machine. It was about him, too. Oversight was necessary, like a captain steering through a storm. He couldn’t rely solely on Orion; he had to use his own skills, his own intuition. There were no established routes in this sea, no charts to guide him. He had to make his own way.

“Testing frameworks for AI-generated code are like uncharted waters,” Santiago thought. “We’re pioneers in this sea, and we have to develop our own methods for ensuring we stay on course.”

He took a deep breath and dived back into the code. He started focusing on the more complex issues, the big fish that Orion couldn’t catch. “These are the challenges that require human intuition,” he realized.

And when he did use the AI, he did so with caution, like a fisherman testing unfamiliar waters. “You have to know when to rely on the machine and when to trust your own instincts,” he thought.

As he worked, he realized that this was the reality of the sea of code. Man and machine, both flawed, both learning. It was a battle and a brotherhood, a journey with no end. But it was his journey, and that made all the difference.

In the dim light of his room, with the sea of code stretching out before him, Santiago felt like a true fisherman of the digital age. And for the first time, he didn’t feel alone. Orion was there, too, a silent companion in this endless voyage.

In the end, I find, we are all like Santiago, facing the sea alone with our tools. The machine, like Orion, is there, but it’s not the whole story. The sea is big, and we are small, but we go out anyway.