The best AI for habit forming is built on psychology, not streaks

Four findings from behaviour-change research do most of the work. First, a plan for the moment beats a goal. "Go to the gym more" fails; an implementation intention, an if-then plan made in advance such as "if it is 7am on a weekday, I put on my shoes and go", reliably helps. A 2006 meta-analysis by Gollwitzer and Sheeran across 94 studies found a medium-to-large effect on reaching goals.

Second, self-monitoring. Recording whether you did the thing changes whether you do the thing; a 2016 review in Psychological Bulletin by Harkin and colleagues found that monitoring progress makes goals more likely to be reached, especially when the record is made physically or shared.

Third, recovery after a miss. In a 2010 study of habit formation, Lally and colleagues at University College London found that missing a single opportunity did not materially affect how a habit formed. What hurts is the interpretation: one lapse, read as failure, becomes permission to quit. Relapse researchers call this the abstinence violation effect.

Fourth, review. Habits rarely fail at random. They fail on the same day, at the same time, after the same trigger, and looking back across a week is how the pattern shows up.

Ari's habit coaching is built on behaviour-change techniques, not on streak counts alone. When you set up a habit, Ari helps turn "I should" into a concrete "when X happens, I do Y" plan, sized for your worst normal day rather than your best one. Ari also uses WOOP (wish, outcome, obstacle, plan) to name the obstacle before the plan, and habit stacking to attach a small action to something you already do.

Self-monitoring becomes a text message. Ari checks in at the time the habit needs it, so the record does not depend on you remembering to open an app, and you answer the way you answer a friend.

Recovery is the core of the design. Ari coaches by one rule: never miss twice. After a miss the next step gets smaller and sooner, never a lecture. After two misses in a row, Ari asks what is in the way and whether the habit is the wrong size, instead of adding pressure. A missed day resets the current streak, your best streak stays on record, and rest days you plan in advance do not break a streak.

Once a week Ari reviews how each habit went and suggests at most three small adjustments, then checks at the next review whether each change helped and keeps it or rolls it back. Weekly reviews and scheduled check-ins are part of the paid plan.

A habit slipping is usually about something else: a work trip, a bad week of sleep, a new shift pattern. Ari's memory is a temporal knowledge graph built on Graphiti, so the facts of your life are stored with dates, and a daily review of your recent conversations records behaviour patterns with the evidence behind them.

That is what lets Ari say "Thursdays keep slipping since the late shifts started" instead of "you missed a day". You can open the memory explorer to see what Ari remembers, and correct anything by telling Ari.

A habit tracker waits for you to open it. Ari comes to you: check-ins arrive as messages on iMessage, Telegram, or Discord, in the coaching style you choose, from hype to quiet support.

If you go quiet, Ari sends one light line instead of piling on. Ari can also build you a small app for a habit, such as a log for a training block with a streak chart, and with Google Calendar connected, Ari can plan a habit around what is already on your day.

Ari is AI, not a therapist or a human coach, and can be wrong. For anything touching mental health, a person is the right help.