The Heart Rate Monitor Experiment: When AI Agents Become a Drug
· 6 min read
Most software engineers have, at some point, flirted with flow state: the intoxicating buzz of deep concentration, where the world fades and the code becomes all. And on paper, agentic coding should be a cheat code for that state: multiple AI copilots handling boilerplate, offloading context, and helping you stay in the zone. More output, fewer distractions. A deeper kind of flow.
But this was something else.
I strapped a heart rate monitor on myself not to train for a marathon, but to answer a quieter, weirder question: what’s actually happening to my body when I go full Agentic Mode?
Spoiler: it wasn’t Zen. It was Vegas.
The Experiment: Code, Agents, and Heart Palpitations #
I strapped on a heart rate monitor while coding, just to see. At rest, I’m a cool 69 bpm, meditatively alive, slightly caffeinated. Then I initiated what I call “agentic mode,” which is just a fancy way of saying “me + multiple AI copilots + too many open projects.” With two concurrent projects, my heart rate jumped to 95. With four, it hit 110.
At that point, my partner asked a perfectly reasonable question. My internal response was not reasonable. I clenched my jaw like a stressed-out raccoon and stared into the existential void between loving someone and wanting to throw your laptop across the room because they interrupted a particularly spicy line of YAML.
It wasn’t just irritation. It was limbic-system-fueled fury. The kind of irrational defensiveness you usually reserve for git conflicts or disappearing breakpoints. And it was completely out of proportion.
Which is when it hit me: I wasn’t just coding. I was stimulating my sympathetic nervous system like a caffeine-soaked day trader.
The Physiology of Overclocked Focus #
The data was unmistakable. My body was treating every additional AI agent like a shot of espresso straight to the nervous system. But here’s what bothered me: I felt more productive. My mind was a buzzing with focused attention, even as my heart rate climbed toward cardio workout territory.
This wasn’t Csikszentmihalyi’s flow1. That kind of flow feels timeless and effortless. This was flow’s hyperactive twin: everything everywhere all at once, but with a tight chest.
Heart rate variability (HRV)-a marker of autonomic nervous system flexibility-tends to decrease as cognitive load increases. Studies show that mental workload alters cardiovascular function, lowering the variability that indicates a healthy, adaptable stress response2. Translation: your body starts preparing for war, even if the enemy is just a segfault.
I could feel it. As I added more agents and projects, I felt smarter, faster, invincible. But also wired, distracted, and irritable. My mind was a cathedral. My nervous system was a rave at 3am.
The Crash #
The moment partner interrupted, I wasn’t just pulled out of flow. I was ejected from orbit, re-entering the atmosphere with all the grace of a burning satellite. The emotional whiplash? Real. And it lingered. Research confirms that developers self-report “spikes in annoyance, elevated heart rates, and residual resentment even after brief context interruptions”3. Yes. That. All of that.
I should clarify: I love my partner. But in that moment, she wasn’t a person. She was an alert box in my mental IDE. And I wanted to click “X.”
This is the dirty secret of our AI-assisted productivity renaissance: we’re not just optimizing our work-we’re optimizing our nervous systems into a state of chronic arousal that masquerades as peak performance.
The Addiction Cycle #
Here’s what I think is happening: We get bored. The single-threaded work loses its challenge. So we add an AI agent, then another, then another, each one promising to unlock a new level of productivity. And it works! For a while. We feel like productivity wizards, orchestrating a symphony of parallel processes.
But we’re not actually multiplying our cognitive capacity-we’re fragmenting it. And our bodies are paying the price.
Cognitive psychology has already laid down the warning signs: every additional context increases the cost of switching exponentially4. What I was doing was more akin to rapidly shifting between identities: Which version of you was in charge of the payment pipeline? The logging refactor? The rabbit hole of BFF endpoint mediation? Forget a tab, lose a self. We are not designed to context-hop with grace. We are designed to forage, focus, and nap.
Each interruption didn’t just cost time. It cost self-reassembly.
The Wake-Up Call #
More seriously, the physiological impact of this kind of cognitive juggling is hard to ignore. We joke about burnout, but the data on sustained elevated heart rate and diminished HRV isn’t cute. It’s clinical. Chronic stress, even the ‘good’ kind disguised as flow, wears down the heart. The body treats every Slack ping like a predator. And if you repeat that pattern long enough, it stops being metaphor. It becomes hypertension, poor sleep, and eventually a visit to a cardiologist who probably doesn’t want to hear about your git rebase strategy. Although if they did, I imagine they’d prescribe fewer interactive rebases and more interactive naps.
Finding Balance #
Flow isn’t evil. But chasing flow through aggressive multitasking and agent overuse is like flying with a Red Bull IV drip. You’ll get somewhere faster, sure. But maybe not where you wanted to go.
This experience has made me reconsider how I structure my time. Maybe running four simultaneous projects with AI agents isn’t the productivity superpower I thought it was. Maybe it’s performance art in the genre of stress management failure.
Now, when I go into agentic mode, I set limits. I log off earlier. And when my partner interrupts me, I try. Try to see them as a person and not a CPU spike.
AI agents are powerful tools. But like any powerful tool, they can be misused. The question isn’t whether we should use them-it’s whether we can use them without sacrificing our health, our relationships, and our humanity in the process.
Sometimes, the most productive thing you can do is write one good function. Slowly. With your heart rate under 80.
-
Csikszentmihalyi, M. (1975). Beyond Boredom and Anxiety. Jossey-Bass. https://scholar.google.com/scholar_lookup?title=Beyond%20Boredom%20and%20Anxiety.&author=M.%20Csikszentmihalyi&publication_year=1975& ↩︎
-
Delliaux, S., Delaforge, A., Deharo, J. C., & Chaumet, G. (2019). Mental Workload Alters Heart Rate Variability, Lowering Non-linear Dynamics. Frontiers in Physiology, 10. https://doi.org/10.3389/fphys.2019.00565 ↩︎
-
Shakeri Hossein Abad, Z., Noaeen, M., Zowghi, D., Far, B. H., & Barker, K. (2018). Two Sides of the Same Coin: Software Developers’ Perceptions of Task Switching and Task Interruption. ACM. https://dl.acm.org/doi/10.1145/3210459.3214170 ↩︎
-
Borst, J. P., Taatgen, N. A., & van Rijn, H. (2010). The problem state: a cognitive bottleneck in multitasking. Journal of Experimental Psychology: Learning, Memory, and Cognition, 36(2), 363-382. https://pubmed.ncbi.nlm.nih.gov/20192536/ ↩︎