lo & slo: a summer series

starting something this week. a clarai series called lo & slo, 5 essays across the summer about a single gap.

the modern ai stack moves at a speed human biology was never built for. and the people getting the most out of it are quietly the most fried.

if you’ve felt the version of it in your own week, the laptop you can’t quite close, the workload that keeps expanding, the strange experience of being more productive and more exhausted at the same time, you’re not doing it wrong. the system is.

3 studies from the last 90 days, by serious researchers in serious venues, confirm the gap is real and measurable.

uc berkeley haas¹ (HBR, feb 2026): aruna ranganathan and xingqi maggie ye spent 8 months inside a 200-person tech company, interviewing and observing staff using ai daily. HBR runs it as in-progress research. 83% said ai increased their workload. exposed workers put in an extra 3 hours 15 minutes per week on average. the researchers named the dynamic the ai burnout paradox, workers voluntarily expanded their workloads because the tool made more feel achievable. by month 6, exhaustion and decision paralysis set in.

bcg (HBR, march 2026)²: julie bedard and colleagues surveyed 1,488 us employees at large companies. acute cognitive fatigue concentrated in the heaviest ai users, especially those managing several agents at once. mental fog. headaches. slower decisions. they called it “brain fry.” high performers got hit hardest.

mit / nber (february 2026)³: daron acemoglu, an mit economist who won the 2024 nobel memorial prize, and 2 coauthors built a formal economic model of how ai-driven cognitive offloading erodes society’s shared “knowledge commons.” their conclusion is bracing: at a high enough level of ai accuracy, the stable steady-state is what they call complete knowledge collapse.

3 different research methods. 3 different timescales. the same arrow.

what’s interesting is that a popular frame for what’s underneath the data isn’t new. stanford psychiatrist anna lembke⁴, chief of stanford’s addiction medicine clinic, laid it out in her 2021 book dopamine nation. she calls it the pleasure-pain balance: pleasure and pain are processed in the same brain region, and they oppose each other. every dopamine hit creates a counterweight of low-grade pain. the only short-term relief is another hit. it’s pop psychiatry more than rigorous neuroscience, but the shape matches the architecture behind slot machines, social feeds, junk food, doomscrolling, and what knowledge workers report feeling at the end of an ai-heavy day.

here’s where ai is different, and where this matters for knowledge work specifically. ai doesn’t deliver pleasure the way instagram or potato chips do. it delivers something far more potent for people who do thinking work: the feeling of being productive. a draft in seconds. an answer in seconds. a deliverable in minutes. 10 variations of a strategy memo before lunch. each one a real, small accomplishment. each one a real, small dopamine kick from a tool nobody warned you was addictive.

so you keep prompting. you can’t quite close the laptop because closing it means losing the next hit. and the pain underneath, that productive-and-exhausted feeling, keeps growing. it’s like an ai hangover, a chronic cycle of restlessness when the screen is off and mental fog when it’s on.

that’s the gap, and the whole series in one paragraph: every dopamine hit your ai stack delivers is borrowed against your future cognition. when a biology evolved for scarcity meets an engineered system that never stops, the brain pays for that friction in measurable ways.

the debt is paid in acute burnout, knowledge collapse, and measurable cognitive decline. each essay in this series walks through one of those bills, and one angle on how to think about it.

the lo & slo series begins and ends with a challenge: we close the laptop on weekends.