Working with AI is exhausting.
Not because it’s hard. Because it’s endless. Every answer opens three more questions. Every prototype suggests two more variations. Every task you complete reveals something else that’s now suddenly within reach. The pace is relentless, the turnaround immediate, and the list of things you could do grows faster than the list of things you have done.
I’ve written before about the importance of protecting time for thinking rather than just doing more. That instinct has only intensified. The volume of what’s possible—the constant iteration, the ability to move from idea to working thing in hours—creates a kind of cognitive pressure that’s hard to articulate. It’s not burnout exactly. It’s the weight of permanent opportunity.
The Analog Response
I’ve been reading Cal Newport’s Slow Productivity and finding myself drawn to its central argument: do fewer things, work at a natural pace, obsess over quality. One of his key moves is cutting out the small—eliminating the overhead, the administrative debris, the low-value tasks that fragment your attention. AI should be perfect for this. It can handle the small so you don’t have to. But in practice, AI often does the opposite: it generates more small. More options, more threads, more things to review and react to. The tool that should free your attention ends up consuming it. (Sounds like computers, email, and Slack all over again.)
Ideas need time to breathe. They need to arrive half-formed and sit for a while. They need to collide with other half-formed thoughts before they’re worth pursuing. So I’ve been reaching for analog forms. Walking without headphones. Notebooks. And increasingly, voice.
Speaking Instead of Typing
For the record: this is not about whispering to the AI. Voice-as-input for AI has its own merits and is widely popular with developers. This is something different.
Voice recording is new for me as a thinking tool. I’ve always been a writer—capture everything in text, process it, turn it into something. But I’ve noticed that the act of typing pulls me into editing mode. I start structuring before the thought is fully formed. The tool imposes its own pace. (Even seeing words transcribed in real time has the same effect.)
Speaking is different. When I record a thought, I’m not organizing it. I’m not formatting it. I’m just externalizing it—getting it out of my head and into the world before my internal editor can intervene. The capture happens at the speed of thinking, not the speed of processing.
This matters more than I expected.
The Gap Between Capture and Processing
There’s a tension here. I want my ideas captured and fully processed, but without requiring my presence or attention.
I want voice notes transcribed. I want them searchable. I want them available to AI when I’m ready to work with them—as context for a broader knowledge base, as triggers for workflows, or simply as a searchable archive of thinking. All of that should happen. But it shouldn’t need me there watching it. The moment of capture should be frictionless and human-paced. Everything after that is the machine’s job.
The immediate feedback loop that makes AI so powerful is exactly what makes it exhausting. You say something and it responds. You iterate. You refine. And suddenly you’re an hour deep into a direction that started as a passing thought—one that deserved to remain a passing thought for a while longer.
Async by Design
What I’m describing is an intentionally async relationship with AI. Capture now. Let the machine process on its own time. Think at human speed, and let AI do its thing—but on your terms, not its.
I’m not rejecting AI. I’m refusing to let it set the pace. The capture layer should be calm, immediate, low-friction. No screens if possible. No feedback loops. Just: record the signal.
The processing layer can be as sophisticated as it needs to be. Transcription. Summarization. Organization. Integration with the tools you already use. All of it can happen—just without demanding your presence or attention. You don’t need to be in the loop for the machine to do its work.
The Discipline of Space
Of all the things you can now do, which do you choose to do? That question still sits at the center of working with AI. But there’s a prior question that doesn’t get enough attention: when do you choose to do them?
Not every captured idea needs to become a task. Not every thought needs immediate follow-up. Some of the best ideas are the ones that sit untouched for days and are still compelling when you return to them.
The discipline isn’t in doing more with AI. It’s in preserving the space where ideas form without AI demanding your attention. I’m setting up my own workflows so I can trust that the infrastructure works whether I’m watching or not—that nothing is lost by stepping away from the screen while the back end does what it does.
The hard part isn’t the technology. It’s embracing—and demanding—that AI does its job without you supervising every step. Until you’re ready to collaborate.