Touch Grass
Your AI agents don’t need breaks, but you do
What does snorting cocaine and working with AI have in common? The non-stop rush. After a week in Virginia, talking with students, faculty, and industry professionals, it felt like I’d traveled back in time to Miami in the 80s. There was the normal population behaving, well, normal. And then there were these individuals that were completely racing, all the time.
Obviously, the fundamental flaw in this comparison is that there is no upside to cocaine, but there is plenty of incredible benefit to AI usage. That is why people who get a taste of this new wonder drug are so quickly hooked. Your productivity accelerates astronomically. And the more time you spend tuning your work with AI, the more productive you get. When you build an app in one day that would’ve taken weeks before, you get a huge dopamine hit. And when you discover how to effectively use an agent to help drive your workflow, you get another hit. Articulate your playbook for triaging bugs, and you get another hit.
Your Energy Flywheel1 is spinning faster with every AI hit you take. In all of my workout mentality focus on revving your Energy Flywheel, I never thought there was such a thing as running your flywheel too hot. But AI is like adding nitrous oxide into the system. It’s dangerously fast. Unchecked, your AI-accelerated flywheel can spin so fast that it’s stressing the infrastructure that’s holding the flywheel in place. The centripetal stress is threatening to tear the wheel itself apart. The bearings are reaching their functional limits. One slight change in any part of the system can lead to a catastrophe.
You can’t run nitrous through your engine continuously. The engine needs to cool down. So too do you need breaks from AI-injection to keep your flywheel sane.
No manager or colleague is going to talk you out of taking more AI hits. You’re all together at the “party”. You have to self-regulate, for your long term sanity.
Here are some key comments from my conversations last week that raise warning flags for me. Have you found yourself saying any (or perhaps all) of these? We’ll dive into each in more detail below.
“I feel guilty not working when my AI is.”
“I’m simultaneously exhilarated and exhausted.”
“I’m learning a ton, but I feel like I don’t know anything.”
Race against the machine
“I feel guilty not working when my AI is.”
Feeling guilty about your contribution level to your team’s output is a pretty typical experience. You’re working with your colleagues and noticing day after day that they all seem to be working more intensely than you. It’s not that they’re staying later than you or taking shorter lunch breaks, it’s just that when they are sitting at their desks working, they seem to be focusing more than you are and moving through their work faster.
This is something that I experienced in my first open space working environment, and it actually helped my productivity. My colleagues were all modeling this behavior, so I knew it was possible. And I found a renewed focus when I felt the pressure of letting my peers down.
Human to human, peer pressure can be a positive. But when that peer is a machine, you need to keep in mind the biological and physiological differences at play. It’s a losing proposition to try to keep up with a machine that never needs to eat, drink, rest, or sleep.
Set more humanly realistic expectations, using a Play the Long Game2 mindset that ensures your routines are sustainable. To avoid “losing the plot entirely”3, put checks in your system. One professional I talked with last week has a “Touch Grass” status on his Slack, to let his team know he was outside taking it all in. In these breaks, remind yourself of how much more productive you are, and congratulate yourself for being productive enough.
You’re never going to outpace AI, but you will always outdream AI. Remind yourself that your uniqueness is what is most valuable, and give yourself the needed down time for actual dreaming.
“Don’t stop me now …”
“… Because I’m having a good time, having a good time.”4
When I hear, “I’m simultaneously exhilarated and exhausted,” it reminds me of my first experience managing a multi-geo team. Most of my feature teams were in Washington, but I had two feature teams in India, 13 hours ahead of us. We would crank all day in Seattle, and then while we were sleeping our colleagues in Hyderabad were cranking away. Any given day, the volume of visible progress was exhilarating.
When my team was all local in Seattle, there were a few members that would end up staying later at work. As the early riser, I would arrive at work the morning after with a manageable amount of work to get up to speed on what had progressed the evening before. I would quickly catch up and then start setting things up for my team for the day ahead.
With my multi-geo team, I arrived in the morning to find 10 people’s worth of progress (and questions and follow-ups) that I needed to catch up on. I struggled to contain my morning catch-up routine, and eventually tried to solve this by bringing my laptop home at night and checking mail before I went to bed. It was exhausting.
The solution was to uplevel my approach to the management of my team such that I wasn’t robbing energy from tomorrow to get through today. That same upleveling mindset can help make your partnership with AI more sustainable.
Where is frustration highest with your AI agents? What additional guidance can you give the agent that will lessen this frustration? What new agents can you introduce to offload some of your duties? If the volume of work that the agent is generating is becoming overwhelming for you to review, construct a reviewing agent and evolve it to the point where it produces an accurate and succinct report for you to digest.
The “AI accelerated workflow” space is still in its infancy. The downside is that there aren’t yet many hard and fast best practices. But the good news is that new ideas are coming constantly. Keep your eyes and ears open for them. Brainstorming with colleagues about their AI flows is a great way to riff off of each other. Make it a conversation over a team lunch and you’ll be tuning your system while you’re touching grass.
Let the learning sink deep
AI accelerates learning and makes it more engaging. AI can tailor instruction to your exact base of knowledge to ensure that you are efficiently learning new material in a way that relates to your own frame of reference. And the fact that the generated lesson plan is uniquely yours will increase your own interest level in the learning experience.
In the movie The Matrix, you see a futuristic learning ideal, where the computer feeds you new knowledge via a direct neural link. In seconds of transfer, Neo wakes up and declares, “I know Kung Fu.”
AI isn’t there yet, but the dopamine rush from the current just-in-time, tailored-to-me education is already more than enough to leave you hungry for more. It’s easy to just learn one more thing. In The Matrix, once Neo experiences this lightning fast learning, he can’t stop. Tank tells Morpheus, “Ten hours straight. He’s a machine.”
“I’m learning a ton, but I feel like I don’t know anything,” captures the gap between The Matrix ideal and the AI reality. AI is helping to present the information to us in a way that eases absorption. But how do we make sure it sticks? There have been a number of studies on AI’s impact on “durable knowledge” and “deeper semantic processing,” warning of the potential for AI to create “cognitive offloading and superficial learning.”5
In 2011, well before today’s AI boom, there was a study on the “Google effect”6, where convenient 24/7 access to searching the vast internet led people to rely less on independent retrieval of information. “Why do I need to remember important dates from the Civil War when I can just ask Google whenever the need arises?” There is a lot of truth to this, but if there isn’t enough knowledge in your brain for independent retrieval, we create an overreliance on the computer.
Spontaneous thoughts are the gems delivered to you compliments of your own Default Mode Network (DMN)7. Your DMN has no access to Google (well, not until Neuralink pulls this off). Your brain, independently and marginally directionally, is constantly processing and synthesizing all of the knowledge it has accumulated. This leads to novel thought where you can experience “A-ha!” moments, big and small.
Give yourself down time between each learning experience. In 15 minutes, you can create very meaningful space for learning reinforcement. Here’s the full progression of the Learning Reflection habit.
Crawl: Record (2 minutes) - when you’ve completed a learning exchange with AI, create a page in your External Brain8 and copy all the relevant links and dialogue to that page.
Walk: Summarize (3 minutes) - add a short summary of that learning experience to the top of the page. Don’t let AI write that summary for you!
Run: Reflect (5 minutes) - Re-read the entire page once more and update the summary with anything you missed. Then set a 2min timer, close your eyes, and just think about that.
Fly: Uplevel your Learning (5 minutes) - Pop the stack and reflect not on the material you learned, but on the learning experience itself. Did you become a better learner? What new learning device or strategy did you discover? Add these insights to a “How I Learn” page or section in your External Brain.
Probably no surprise that the learning content at the end resonated highly with me. One step I would add into all of the fantastic refection is some form of *application* where you take action on the learning, and then add that experience to your reflections!
Great point on "taking action on the learning", Chris. I can always count on you to up the learning game.