@nolan there's quite a bit of experimental research around that idea these days (I'm biased, it crosses my creativity research); even Anthropic released an early paper that shows skill formation and mastery degrades with use https://arxiv.org/pdf/2601.20245 —though methodologically it could use some work.
Discussion
Loading...
> There used to be a tax on ideas, and it just hit zero.
Perhaps the tax is gone, if your Claude bill is just an incidental, but your supplier dependence has just gone through the roof. This is putting aside the material costs of this form of computation, and the epistemic debt it incurs. My prediction: we will look back on this type of willful ignorance with horror.
@nolan FWIW I've started to scale back parallel agents and build in enforced breaks during my day. I found myself on the path to burn out otherwise. Maybe that will make me less competitive, but it seems better to find the right balance between productivity and sustainable pace.
@ben Yeah I've considered doing "no-agents Thursdays" or something as well. It's extremely exhausting otherwise.
@nolan I limited myself to at most 2 parallel agents (tasks I was involved with) and let myself play some video games over my lunch hour. Felt much better at the end of the day and I think I still accomplished quite a lot.
Honestly maybe this is where a lot of the "agent psychosis" comes from. I used to be a very strict 9-to-5er, but it starts to look at bit silly when your contribution to the development process is a quick gut-check on each agent's 5-20 minute iteration loop. I can imagine people just fielding agent questions all day long, the way you might field Slack messages. I'm resisting, because I value my free time too much, but I can imagine getting sucked in.
Of course all of this is also the exact opposite of "flow," and very likely to degrade the very skills that you're exercising when you point out stupid mistakes to these robots all day long. Although you could also argue that this is what a senior manager or director does – trade in their hard-earned skills for guidance to more junior devs, while the muscle they built up in the trenches starts to slowly atrophy. Maybe the difference is just that more people are "cashing in their chips" now.
@nolan there's quite a bit of experimental research around that idea these days (I'm biased, it crosses my creativity research); even Anthropic released an early paper that shows skill formation and mastery degrades with use https://arxiv.org/pdf/2601.20245 —though methodologically it could use some work.
@justin Yeah this seems unsurprising, similar to how I got worse at bar trivia after I had Google in my pocket. 😆 What's interesting is it seems there's the potential for many more kinds of skills to be degraded, since you can use it as a prosthesis for so many more things.
@nolan intuition is a hell of a thing 😂 it tells us it's unsurprising given lived experience and then we spend a lot of time to find the edges research-wise (be it in the wild product development or structured experiments) to see where potentially we find the value or efficiency. But people feel it across skill sets even if they haven't seen it yet in their field—to ignore it, to not consider or think about AI and its effect, is not a workable position in my mind.