Work at Race Pace_

The eight-hour workday made sense for a factory. The machine kept running, so people stayed beside it for the duration of a shift. Technology changed the work but we kept the shift. Even remote companies often hire people as if they are buying access to the better part of the day. Slack moved the factory floor onto a screen, but the expectation survived. You are meant to be around!

The popular story says AI will help us work less. I am not convinced. My experience is that it compresses the work and leaves us tempted to fill every minute we saved. I have already written about becoming an AI vampire when the model kept answering fast enough to make stopping feel stupid. That was mainly about working longer. The deeper change is that we can now work much harder for a much shorter period.

I learnt the shape of this years ago by playing StarCraft... badly. When I first started I loved the build-up. I wanted to finish my economy and upgrades before building the perfect army. Then I would have one glorious final battle. This felt very satisfying. It was the wrong way to play the game and a lovely way to lose.

Real-time strategy does not let you finish one part of the game before the next begins. The economy is moving while units are being produced. The battlefield changes while you are deciding what to build. Everything competes for the same attention. A match can be over in 15 minutes and still leave you mentally finished.

AI work has the same choice hiding inside it. You can play it slowly or you can play it like StarCraft.

You can give one agent a task and wait for it to finish. That spreads the work through the day and keeps the pace manageable. The intense mode begins when you start a second session instead of waiting. Then a third. Soon one agent needs a decision while another returns work that needs verification. You have to remember why both tasks exist and where they fit into the larger thing. The job starts to resemble managing a live system.

I have tried doing this with many agents at once. I do not do it every day because it is exhausting. The agents can keep producing long after my ability to connect their work starts to weaken. At that point I am the connective tissue across dozens of moving parts. In theory the system can grow to hundreds of sessions. In practice the human coordinating them becomes the limit much sooner.

Formula One has a similar constraint. A race may last around two hours, yet a driver who finishes one has already absorbed the physiological load of sustained competition. Intensity changes what an hour means. AI is beginning to do the same to knowledge work.

This is also why the fried brain of cyberpunk no longer feels so ridiculous to me. I used to imagine some unsafe implant cooking your brain with electric shocks. The ordinary version already exists in burnout. Nothing has to physically burn. A person only has to sustain competition-level focus for longer than they can recover from it. AI raises the risk by concentrating more consequential decisions into every hour.

The obvious mistake is keeping the eight-hour expectation and adding AI acceleration on top. That turns every productivity gain into more work for the same person. A four-hour day of intense agentic work may carry the load of a conventional day. The remaining hours are part of the recovery cycle. Treating them as spare capacity removes the recovery that makes the pace possible.