Be Egon

To the Point

Consider AI tooling an input‑sensitive approximator: given clear constraints, context, and examples, it tends to approximate the output you want. Start by examining your input before blaming the tool when results disappoint.

“Garbage in, garbage out.” More usefully: sharper in, sharper out. Your job is to iteratively approach a sharper input.

Use AI to help craft that input. Decompose the problem and refine step by step.

With Humor

A moment from 1984’s Ghostbusters captures this posture.

The movie stars three scientists at Columbia University, researching the paranormal: Egon (pragmatic and tool‑centric), Ray (more of a dreamer than Egon, but similar), and Peter (the glib showman — he’s in it for the paycheck and respects science less than the others). Not an unlikely crew, I’d say.

After years of nothing, they finally stumble upon a real ghost at the New York Public Library in an iconic opening that brings their research to the forefront.

They go into business for themselves. Their first job? A green floating blob fans would come to know as “Slimer” in a luxury hotel not far from the library. And here’s the moment.

The crew assembles in an elevator, geared up with complex machinery we learn is capable of more or less total destruction. In that moment they have a startling realization.

Ray: You know, it just occurred to me that we really haven’t had a successful test of this equipment.
Egon: I blame myself.
Peter: So do I.

While on paper that reads as though Peter also blames himself, I’ve always taken it to be him blaming Egon. A debate for Comic‑Con, I suppose. That’s not what we’re here for though. We’re here for Egon.

In that moment, chasing a solution to a problem that arguably only existed in their minds until that infamous day in the library, things were not perfect. There are many reasons why — but immediately our hero chimes in: I blame myself.

In the movie they just keep moving, of course. Everything goes well. There’s a giant marshmallow man that attacks New York City — but ultimately this previously unproven technology prevails.

We don’t care about that part of the story. We’re here for the courage to point at yourself when things aren’t perfect. You can always blame someone (or something) else. “It’s a poor carpenter who blames his tools,” as they say.

How to Be Egon (in practice)

In agentic software development, take this lesson to your inputs. What did you give the agent?

And when the agent fails?

I blame myself.