This is naturally a part of any complex work, but also is an integral part of creating systems.

Please note, this is only part of creating systems, not using them. Much needed to keep the end goal in mind and not get lost in details.

But, here's the interesting twist:

Fresh eyes should be a part of any process anyway. If you don't build it in, the output will suffer. Seems obvious, but it always surprises me how often it gets missed. Then someone will realize this and quietly slip a break in.

In fact, if done just right, it dramatically increased the quality of output and in many cases the overall speed. It's like shifting gears in a car at the sweet spot, and not when the RPMs are screaming.

Funnily, this also includes LLMs to a certain extent. This is not just about the context of your chat, the issue persists even after starting a new chat or clearing context. You can see the output suffers when there is too much load or traffic for that AI model in general. You literally stop and do the same things after sufficient gap and it starts working like it used to.