The Effects of GenAI and Slacking on Software Development

Anecdote: I developed a custom driver for an environmental sensor three years ago. I got a sample of a newer, better variant of the sensor last year, but I hadn't found the time to update the driver. A few days ago I turned it over to a Codex 5.3 agent. Not surprising to those keeping up, but the agent did it. As in, it did a fantastic job converting the original repo to a new, working driver; updating all code, references, and names; addressing some interesting nuances in the driver; and adding support for a new feature. What would have taken me much of a day was replaced by 1 prompt and clicking "Allow" several times (could've safely done Auto Approve). By waiting a year to do the update I reduced my time from 4-8 hours, to ~10 minutes (plus an hour for testing; review; and clean-up of a URL and some formatting). The current generation of AI easily did an excellent job with this modest project.

I'm reminded of the facetious paper "The Effects of Moore's Law and Slacking on Large Computations", where the authors showed that, due to the expected speed gains in the successive generations of hardware, one should wait to start some very arduous computations:

... you could go to the beach for 2 years, then come back and buy a new computer and compute for a year, and get the same amount of work done.

So, there's a point where:

The optimal time to begin any more arduous computation is in the future (after an optimal amount of slack time).

There are plenty of reports of AI's limits on many complex projects today, but in such cases I'm wondering: Is the optimal time to do some arduous software development "in the future" (when AI is even more proficient than it is today)?