Agentic Engineering Frees Your Working Memory

One massive advantage of agentic engineering is you no longer need to hold the entire mental model of your codebase in your brain’s working memory.

This makes it much easier to stop and start work.

Previously, hand writing code typically meant “loading” all of the context of your project into your brain and then writing code with a degree of flow state. The problem was, interruptions or days off away from the code l, and that working memory evaporated. Forcing enormous friction when you sat down to begin again.

This proved incredibly frustrating and tiring. Patience for the rest of the world diminishes.

The pre-ai antidote to this ended up being prolonged coding sessions, phone in a drawer (or another room) and generally disconnecting from reality to fully give in to the required flow state, while attempting to retain as much code context in my head as possible.

Post December, this changed.

Prompting is so much more ephemeral. I have an idea and a prompt it into existence; the agent holds the context and builds the mental model of my project.

It’s liberating. It’s made me re-think how I structure my day, how I approach tasks, how much I can context switch, and, importantly, how much easier it is to switch off at the end of the day.

I still strongly retain the architectural model and scaffolding in my head to ensure I’m guiding the ai. But I’m sure this will fade over time as trust builds, and accuracy improves.

We’re all still learning this new way of working. So far, it’s a welcome change.