Every session in this room is built from five mechanisms with published evidence behind them. Each claim below links to its source. Where the evidence is early or mixed, this page says so, because a focus product that overstates its science is just another distraction.
Working on a task in the presence of another person who is simply there. The practice came out of the ADHD community as external support for executive function: another person’s presence seems to supply the activation that starting a task needs.
A 2023 ACM SIGACCESS paper with neurodivergent participants proposed body doubling as a continuum of space, time and mutuality and found it naturally effective for starting and finishing tasks. A 2024 survey of 220 neurodivergent adults in ACM Transactions on Accessible Computing found the same, and that no single configuration suits everyone. A 2025 virtual-reality study of 12 participants with ADHD found they finished tasks faster, and felt more accurate and attentive, with either a human or an AI body double than alone.
What is thin: twelve people is a pilot, not a verdict, and it is the only study that tested an AI double. No large randomized trial has compared virtual with in-person doubling. Objective task completion is mixed across studies; the felt benefit is consistently high.
Zajonc’s 1965 finding: the mere presence of others raises arousal, which improves performance on simple and well-practised tasks. It is the oldest explanation for why a library works better than a kitchen table, and it is why the room keeps a visible count of who is here.
Workplace studies find productivity spillover between co-workers who can see each other working. The mechanism is quiet accountability, not supervision, which is why nobody in this room can see your screen and nobody is called on.
An implementation intention is a plan of the form “when X, I will do Y”. Gollwitzer’s 1999 work, and a meta-analysis of 94 independent tests, found a medium-to-large effect on goal attainment; difficult goals were completed about three times as often when paired with one.
The load phase is that, out loud. “In the next 25 minutes I will write the first section” is a different object from “I should work on the report”. Breaking it into three steps makes the first one small enough to start.
A 2025 scoping review in BMC Medical Education found that time-structured Pomodoro intervals consistently improved focus, reduced mental fatigue and sustained task performance against self-paced breaks. A comparative study found students in long self-regulated sessions reported more fatigue and less motivation than Pomodoro users.
Against that, a 2025 study in Behavioral Sciences found fixed Pomodoro breaks brought fatigue on faster and motivation down faster than self-chosen breaks, with no difference in productivity or flow. The honest reading: structure helps, the right interval varies by person and task. That is why there are four formats and not one.
Sophie Leroy’s 2009 work: when you switch tasks, part of your attention stays with the one you left, and performance on the new one suffers. Her follow-up found that a short written closing ritual, noting where you stopped and what comes next, reduces the residue.
This is the rule behind the room’s silence. Nothing speaks during a sprint, not a chat message, not Hugh. And it is why check-out ends with one written line about where you left off, which is the line Hugh reads back next time.
| Beat | What happens | Why |
|---|---|---|
| Load, 5 min | You declare one intention and can break it into three steps. | Implementation intentions (03). The first step is small enough to start. |
| Sprint, 25 or 50 min | Silence. Hugh is a still tile. The count of people here stays visible. | Body doubling (01), social facilitation (02), attention residue (05). |
| Break, 5 or 10 min | One question: did the step ship? Stand up. | Time-boxing (04). Micro-breaks under ten minutes help; long ones recover. |
| Check-out, 5 min | Done, partial, not started. One line on where you stopped. | Leroy’s closing ritual (05). The line becomes next session’s opening. |
| Next session | Hugh reads your last line back and asks. | Accountability without pressure (02). Nobody is called on; you are remembered. |
That an AI body double works as well as a person: one pilot of twelve people says it might. That this treats ADHD, anxiety or anything else: it is a structured place to work, used alongside whatever care you already have. That the numbers above are ours: they are the researchers’. If a claim on this site ever has no link under it, write to us and we will take it down.
Rule of the house: a citation links, or it does not ship.