Vjeux once told me a story that stuck with me.

It was about the good old days of the World of Warcraft. Blizzard shipped regular updates to the game, but their change logs were very sparse. The community had to figure out the details on their own, and there were a few websites that tried to provide in-depth reviews of the new stuff.

But one of them had an unparalleled level of details. Every little thing was meticulously documented.

People tried to figure out their secret. Did they disassemble the binaries? Did they have someone on the inside leaking the information?

The answer was less magical. The author simply spent hours and days testing the updates manually. It was a lot of labor. Wasn’t it boring? It was. But they were willing to get bored and do the grunt work.

I see parallels to this story everywhere. Progress often comes from someone willing to do boring work. Even writing this blog post required the willingness to get bored editing and rewriting things multiple times.

This is especially insightful today, in the dawn of AI coding. These tools give us so much speed boost and excitement that we are unwilling to slow down.

Everyone I know went through the same experience. The coding agent fails to one-shot the problem in the way its operator expected. The operator either thinks it’s dumb or that they are missing a clever trick to make it work.

But it’s not dumb. And there’s no trick. One just needs to be willing to do the boring work of explaining the context to the model in a durable way.