Most accountability arrangements collapse because both people are too polite to enforce them.
Accountability is the one intervention in this whole field with a decent evidence base behind it, and it is also the one most commonly implemented badly. The standard version is a friend who agrees to check in, means it sincerely, and then does not want to be the person who brings up the thing you clearly did not do. Six weeks later the arrangement has quietly dissolved and neither of you mentions it.
That failure is social, not motivational, which is why these pieces spend more time on the mechanics of the arrangement than on wanting it more. What makes a check-in land. Why external accountability outperforms self-accountability for most people on most tasks, and what the exceptions are. Where an AI companion is genuinely better than a human, which comes down to one narrow advantage: it has no relationship with you to protect, so it can ask the awkward question at the awkward time.
It is worth saying the limit out loud, since the last piece here says it too. Something that never gets tired of asking is useful and it is also not a friend. Those are different things and confusing them causes its own problems.
8 pieces