Thought-provoking article Dan. Great thoughts as always, and I really appreciated your perspective on productivity theater and being wary of pre-mature optimization with these tools.
You mentioned agentic coding tools as helpful for more technical individuals. Do you have any thoughts on the value of agents for non-technical users?
Thank you! I think things like Cowork will be quite useful for a lot of people.
But I think the upside is capped relative to something like Claude Code, because the tasks that Cowork focuses on are less verifiably accurate than code, so they require more and more frequent guidance from humans.
Really thoughtful essay! The execution vs. exploration framing is one of the cleanest articulations I’ve seen. What resonated most: for exploration, the real gain from AI as a thought partner is depth, not necessarily speed. AI doesn’t shrink the time for exploration tasks, it deepens what you get from it. That reframes what productivity even means with AI.
One thing I’d add: how you use AI for an exploration task will also depend on complexity, and where you are in building judgment in a domain. Even within exploration tasks, when your judgment is already strong and complexity is low/moderate, you can lean more on AI to generate while you evaluate. But for other complex exploration tasks, you need to do more of the original thinking yourself and use AI to challenge it.
Agree, I considered adding this point. Basically once you have done enough exploration in a given task/domain and are very confident in it, you can shift it to an execution style approach.
Love this - how do you think about when you should be an early adopter vs when you should wait for folks to figure out ironed out use cases / reliability? I’m feeling this right now with claude code for replacing clay, feels like the future but also a bit too soon :)
I think it is super hard to know. Certainly sooner than in prior waves because of how fast things are changing, but not as fast as X/LinkedIn tries to argue
Love the post and thoughtful explanation of what most people are going through and what steps one can take! I have shared it extensively as it relates to a lot of the conversations that we are currently having!
Thanks for sharing this! I think most people really need this article because it helps us see the big picture. Sometimes, we forget that ‘the way to use’ and ‘why use it’ are more important than just using it.
The pace of change driven by AI has created a state of perpetual high-stakes learning. This demands massive Metacognition—the ability to adapt to new models and IT landscapes overnight. Without training to manage this mental load, workers aren't just "biased" against their limits; they are physically overwhelmed by the sheer speed of the industry.
This training in Metacognition can help in your quest not to go insane !!
Hi Dan, thanks for the note. Yes, metacognition training can help someone (1) be consciously aware of what problem solving strategy they’re using and (2) help them know when and how to change their strategy.
Multi-tasking…tricky business. It is “always” best to single-task whenever possible, it’s where our best, most complete and creative work is done. The trick is to know how to go through the day with a game plan that allows for dedicated focus to each important task. I always screw up if I’m talking on the phone, while typing a thought. Each day I set aside 60-minutes to be cut off from the world entirely when I am working on my emotional recognition AI Agent…close the door, turn the phones off!!
How do you handle it when you’ve got something super important you need to get done? What works for you?
The message in this really similar to my article Decisions and Deliveries and how you can't automate away decisions.
A sort of meta observation: I think us AI builders are all going through a similar experience of excitement ➔ obsession ➔ anxiety ➔ depression ➔ enlightenment when we realise the decision of what to work on stays with us.
Seems like we're arriving at the enlightenment phase now and there's going to be something else next.
Really agree on the dunning kruger effect esp for the learning curve for AI! it takes time and reps until you can really start to understand and therefore make the most out of it. if i were to sum it up, it would be: never stop prompting and challenging it. Thanks for the piece!
Probably the clearest diagnosis of where we are right now. Thanks for putting this together!
Very kind of you, thank you!
Thought-provoking article Dan. Great thoughts as always, and I really appreciated your perspective on productivity theater and being wary of pre-mature optimization with these tools.
You mentioned agentic coding tools as helpful for more technical individuals. Do you have any thoughts on the value of agents for non-technical users?
Thank you! I think things like Cowork will be quite useful for a lot of people.
But I think the upside is capped relative to something like Claude Code, because the tasks that Cowork focuses on are less verifiably accurate than code, so they require more and more frequent guidance from humans.
Como empiezo . A organizar mi carrera artística .
so good! thank you ✨
Thanks Finn!
Really thoughtful essay! The execution vs. exploration framing is one of the cleanest articulations I’ve seen. What resonated most: for exploration, the real gain from AI as a thought partner is depth, not necessarily speed. AI doesn’t shrink the time for exploration tasks, it deepens what you get from it. That reframes what productivity even means with AI.
One thing I’d add: how you use AI for an exploration task will also depend on complexity, and where you are in building judgment in a domain. Even within exploration tasks, when your judgment is already strong and complexity is low/moderate, you can lean more on AI to generate while you evaluate. But for other complex exploration tasks, you need to do more of the original thinking yourself and use AI to challenge it.
Agree, I considered adding this point. Basically once you have done enough exploration in a given task/domain and are very confident in it, you can shift it to an execution style approach.
Love this - how do you think about when you should be an early adopter vs when you should wait for folks to figure out ironed out use cases / reliability? I’m feeling this right now with claude code for replacing clay, feels like the future but also a bit too soon :)
I think it is super hard to know. Certainly sooner than in prior waves because of how fast things are changing, but not as fast as X/LinkedIn tries to argue
Love the post and thoughtful explanation of what most people are going through and what steps one can take! I have shared it extensively as it relates to a lot of the conversations that we are currently having!
Love to hear that. Thank you so much for sharing it!
This is an excellent post and I especially appreciate the attribution to folks like Everett Rogers.
I really appreciate that Beth!
Thanks for sharing this! I think most people really need this article because it helps us see the big picture. Sometimes, we forget that ‘the way to use’ and ‘why use it’ are more important than just using it.
Very much agree. Thank you!
Very grounding piece, thanks Dan!
Thanks Ben!
The "AI Accelerator" and the Upskilling Trap
The pace of change driven by AI has created a state of perpetual high-stakes learning. This demands massive Metacognition—the ability to adapt to new models and IT landscapes overnight. Without training to manage this mental load, workers aren't just "biased" against their limits; they are physically overwhelmed by the sheer speed of the industry.
This training in Metacognition can help in your quest not to go insane !!
Agree - metacognition, and just the ability to effectively multitask - seem to be becoming much more important.
Hi Dan, thanks for the note. Yes, metacognition training can help someone (1) be consciously aware of what problem solving strategy they’re using and (2) help them know when and how to change their strategy.
Multi-tasking…tricky business. It is “always” best to single-task whenever possible, it’s where our best, most complete and creative work is done. The trick is to know how to go through the day with a game plan that allows for dedicated focus to each important task. I always screw up if I’m talking on the phone, while typing a thought. Each day I set aside 60-minutes to be cut off from the world entirely when I am working on my emotional recognition AI Agent…close the door, turn the phones off!!
How do you handle it when you’ve got something super important you need to get done? What works for you?
Really enjoyed this one.
The message in this really similar to my article Decisions and Deliveries and how you can't automate away decisions.
A sort of meta observation: I think us AI builders are all going through a similar experience of excitement ➔ obsession ➔ anxiety ➔ depression ➔ enlightenment when we realise the decision of what to work on stays with us.
Seems like we're arriving at the enlightenment phase now and there's going to be something else next.
I think this will be the main topic of 21. centry.
One underrated takeaway here: AI doesn’t replace thinking, it amplifies structured thinking.
Without a clear system, AI just creates noise faster.
That’s why learning how AI fits into real workflows feels way more useful than just prompt tricks.
This resource goes deeper into that approach if anyone’s exploring it: https://shorturl.at/cvsda
Really agree on the dunning kruger effect esp for the learning curve for AI! it takes time and reps until you can really start to understand and therefore make the most out of it. if i were to sum it up, it would be: never stop prompting and challenging it. Thanks for the piece!
Very much agree. It is too easy to just let it skip you past the painful part!
Cool
Insightful. Different from the usual 'what AI does essays' and talks about :using AI when it is really needed'.
Appreciate that!