During the recent boom of AI Agents, we have the ability to implement our ideas further, and faster than any time in history. It allows the mind to expand its capability to produce ideas since it is no longer throttled by the ability to implement them. With this change, a new bottleneck is introduced, and that is attention, craft, and care. Some ideas will be one-shot, and other ideas should be put off until they can be one-shot, and only the important ideas should be immediately dealt with.

We could imagine a simplified workflow to be something of the sort: [ ] intake ideas into inbox [ ] automated filter from h to l leverage [ ] extract high leverage ideas into a queue [ ] tasks that are either low leverage (entertainment ideas) or likewise too difficult or requiring a significant amount of opportunity cost should be put off.

In retrospect, it's easy to see how well this idea would perform. For projects that could have waited 1-2 years rather than building in the GPT-3 copy-paste, revise era, builders could have saved weeks-months of time by waiting until the current capability of models in the class of 5.5 and Opus 4.7.