The spark fades because dopamine, as Paul Bloom's Psych explains, tracks novelty rather than effort, and hedonic adaptation returns you to baseline within weeks, while the 2024 Singh review of 2,601 people puts median habit formation at 59 to 66 days. Projects die in that gap. The fix is an external cue that arrives without you: Tomo ($19.99 a month, iMessage only) or a scheduled Claude Pro task ($20 a month).
Every founder I coach has a graveyard of week-three projects, and almost all of them diagnose it the same way: "I lack discipline." I used to accept that diagnosis. After a year of running the same experiment on myself with a couple of AI tools and a psychology textbook, I think it is wrong in a specific, fixable way. The spark was never discipline, and its fading is not a character verdict. It is a schedule.
| Phase | What the brain is doing (Bloom) | What usually fails | What the AI check-in should do |
|---|---|---|---|
| Spark, days 1 to 10 | Dopamine fires on novelty and surprise; everything about the project is new | Nothing yet; this is the trap, because effort feels free | Capture the plan and the "why" verbatim while motivation is cheap |
| Trough, days 10 to 60 | Hedonic adaptation: the reward returns to baseline; nothing is surprising any more | The project; you confabulate a reason ("priorities changed") | Message first, at a fixed time, with the smallest next action and no judgment |
| Automatic, day 60 onward | Cue-driven; the median of the Singh review, 59 to 66 days, falls here | Novelty-seeking restarts with a new project | Go quiet; a check-in you no longer need is noise |
The mechanism, from Paul Bloom's Psych. Dopamine is the motivation and reward-seeking chemistry, and what it responds to is novelty and prediction error, the gap between what you expected and what happened. A new project is nothing but prediction error for about ten days, which is why the first week of anything feels like discipline you did not know you had. Then hedonic adaptation does what Bloom says it does to lottery winners and the newly promoted: it returns you to baseline. The work is identical; the chemistry is gone. Bloom's second relevant point is the fundamental attribution error, our habit of explaining behaviour by character when the situation did most of the work. "I lack discipline" is that error applied to yourself. And his third is confabulation: we invent fluent reasons for behaviour we do not understand, so the week-three abandonment arrives dressed as a strategy decision.
The numbers put a date on the trough. Ben Singh and colleagues' 2024 systematic review in Healthcare, covering 20 studies and 2,601 participants, found habit formation reached a median of 59 to 66 days, with means of 106 to 154 days and individual variation from 4 to 335 days. The 21-day figure people quote is, in Singh's words, not evidence-based. So the honest picture is a motivation curve that collapses around day ten and a habit curve that does not carry you until around day sixty. Fifty days of exposed ground. Bloom's behaviourism chapter supplies the bridge: Skinner showed behaviour is held in place by reinforcement schedules, and the cue that reinforces you does not have to come from inside your head.
That is the only thing an AI check-in is for, and I want to be precise, because the category oversells. It is not a coach and it is not insight; it is an external cue that arrives on a schedule you cannot talk yourself out of. Two versions work in October 2026. Tomo, from Mapo Labs, is an accountability assistant that lives in iMessage: you tell it the goal, it remembers, and it texts you first, which is the feature nearly every positive review names. It costs $19.99 a month, it is iOS and iMessage only, several reviewers report that the price appears only after an emotionally warm onboarding, and the small team shows in response lags of 13 to 45 minutes and some downtime. The second version is the one I actually run, because I am on Android half the day: a scheduled task in Claude Pro ($20 a month, or $17 billed annually), which since the Cowork merge can hand off and schedule work. Mine fires at 07:40 with one message, the smallest next action on the project and nothing else, drawn from a plan I pasted in on day two when the "why" was still vivid.
Three rules make the cue work, learned from breaking it. First, the message must be specific and tiny: "open the pricing doc and write one row" survives the trough; "work on pricing" does not. Second, the check-in must be silent on streaks and silent on misses, because a scolding cue becomes a punisher, and Skinner's finding is that punishment suppresses behaviour rather than shaping it; you start avoiding the phone. Third, the cue has a sunset. Once the behaviour runs without it, at roughly the Singh median, I delete the task, because a cue you no longer need trains you to ignore cues.
The limits are the brand here. An external cue cannot rescue a project you chose for its novelty rather than its value; it just makes the abandonment slower. The hedonic curve applies to the tool too: Tomo's texts and my 07:40 message both lose their charge after a few weeks, which is why the message content has to change even when the time does not. And Bloom's deepest warning stands over all of it: your own account of why you stopped is the least reliable evidence you have, so write down what you did, not what you felt about it.
Sources: Ivy, "Honest Tomo AI Reviews (2026): Is the $20/Month Accountability Bot Worth It?," Saner.AI blog, September 16, 2026, pricing, platform and reliability findings (blog.saner.ai); Anthropic, "Plans & Pricing," Pro plan features including hand-off and scheduled tasks, checked October 2, 2026 (claude.com/pricing); University of South Australia via ScienceDaily, "Myth busted: Healthy habits take longer than 21 days to set in," on Singh et al., Healthcare 12(23):2488, 2024 (sciencedaily.com; mdpi.com/2227-9032/12/23/2488); Paul Bloom, Psych: The Story of the Human Mind (2023), chapters on neurons and neurotransmitters, behaviourism, social psychology and happiness.
Related: How to Find Your Passion · Best Self-Improvement Books · How to Make Better Decisions · AI Coach App — Building It in 8 Hours