Coach Memory & Proactive Nudges

Memory and chat history

Coaches use recent conversation and stored memory to maintain context. Chat history is the conversation you can read. Stored memory contains a condensed summary and saved details, such as a name, an accepted nickname, or a lasting preference. It is not a complete transcript, and a coach does not necessarily see every old message when replying.

Summaries update periodically, rather than after every message. A coach can also save an individual memory during a conversation. A reply saying “I'll remember” does not prove that a detail was saved: check the action result and the stored memory.

Coach	Where to open memory	What it belongs to
AI Coach	Open AI Coach settings, then AI memory.	Your AI Coach memory is shared across your sessions; starting another lock does not create a fresh memory.
Bot Coach	Open the coach's chat menu, then Bot Memory.	Memory belongs to you and that coach. Private copies of the same original coach share that memory; unrelated coaches have separate memories.

When between-lock Bot Coach chat is available, its Bot Memory opens the same coach's memory. The nudge diagnostics described below require a Bot Coach attached to a lock.

Review and correct memory

The Memory Snapshot shows Characters, Last Updated, and Messages At Update. These describe the saved snapshot, not the number of messages currently visible or proof that every message was remembered. Stored Memory shows its text, with a usage indicator for your current allowance.

To add or correct a detail, be explicit in chat, for example:

• “Remember that I prefer short check-ins.”
• “Correct the saved nickname: please call me Alex.”
• “Forget the saved preference that I like evening reminders.”

Coaches support adding, replacing, and removing individual saved entries. The memory viewer is read-only; it is not a text editor. Check the result and reopen memory after a correction. Removing an individual entry does not delete mentions of it from chat history or automatically remove every occurrence in the summary.

Memory is context for the model. Storing a preference does not change a permission, cancel a task, or override a lock rule. Use the relevant settings for those changes.

Refresh and capacity

In Bot Memory, Update memory now requests a summary refresh from the available conversation. Manual refreshes are limited to three per hour. A refresh can run in the background; “update started” does not mean a new snapshot has already been saved. Check Last Updated and the text again after it finishes.

A refresh needs an available AI provider and is subject to its usage limits. It can leave memory unchanged when there is no new durable information or no usable conversation to summarize. Ordinary AI Coach settings do not have the same manual-refresh control; its summaries update automatically.

Use the displayed memory allowance as the current limit. If memory is full, additional entries or an oversized import can fail. Removing obsolete entries can make room. A successful chat reply does not guarantee its accompanying memory update succeeded.

Back up, restore, or reset

Export and import

1. Open AI memory or Bot Memory and review the current snapshot.
2. Under Local Backup, select Export to download a JSON file. Export is available when stored memory exists.
3. To restore it later, open the matching coach's memory and select Import.
4. Choose the exported JSON file, then verify the restored text.

An import replaces the stored memory; it does not merge two snapshots. Export the current version first if you may want to recover it. Bot backups identify their coach, and a backup from a different original coach is rejected. Imports must fit your current memory allowance.

The file contains memory, not your full chat history, coach configuration, tasks, or lock state. It is readable JSON and can contain personal details from the conversation, so keep it somewhere appropriate for that information.

Clear memory and clear chat separately

Goal	Control	Effect
Remove AI Coach's stored memory	Clear AI memory in AI Coach settings, or Clear memory in its memory viewer.	Clears the stored AI summary and saved memory; chat remains.
Remove Bot Coach's stored memory	Clear memory in Bot Memory.	Clears memory for you and that coach, including its private copies; chat remains.
Clear AI Coach conversation history	Clear AI chats in AI Coach settings.	Clears conversation history while keeping the initial welcome message and stored memory.
Clear Bot Coach conversation history	Clear Chat History in its chat menu or settings.	Clears the conversation, while stored bot memory remains.

If you want a fresh conversation without the old stored context, clear both history and memory before continuing. Clearing only memory still leaves recent chat available to the coach; AI Coach can summarize retained conversation again. These controls do not reset your lock or cancel existing tasks and actions.

Hardcore Mode

Review memory before enabling Hardcore Mode

The memory controls disable imports and clearing while the applicable Hardcore Mode is active. Clearing chat is also restricted. You can still view and export stored memory, and Bot Memory still offers its manual refresh.

AI Coach Hardcore Mode and Bot Hardcore Mode are separate features. Read the active mode's conditions before enabling it; see AI Coach's Hardcore Mode.

Set up proactive nudges

Proactive Nudges let an enabled coach start a message when you have not replied for a while, including while you are away from the app. They require a coach active on a lock session and an available AI provider. They are separate from ordinary bot actions, task deadlines, and other scheduled events.

AI Coach controls

In AI Coach → Enabled Features, enable Proactive Nudges, review its options, then save:

• Max consecutive nudges: one to five messages before the coach waits for another reply from you.
• Minimum/maximum delay between nudges: the range used to choose the next delay.
• Extra spacing after repeat nudges: keep the same range, or increase spacing by 1.5× or 2× as unanswered nudges accumulate.
• Allow commands during autonomous nudges: enabled by default. Turn this off for messages without autonomous commands. When on, the coach still needs the relevant feature permissions and device consent, and the nudge action limits still apply.
• Quiet hours: choose the hours when nudges should wait, using your account timezone. A range can cross midnight. Setting the start and end to the same hour disables this quiet-hours window.
• Nudge Prompt (optional): guide the tone or purpose of a check-in. Leaving it empty uses the default instructions. A prompt does not grant permission for actions.

AI Coach Hardcore Mode prevents switching off already-enabled Proactive Nudges and locks the Nudge Prompt. Configure these before enabling Hardcore Mode. Timing, quiet-hours, and the autonomous-command switch have their own controls.

Bot Coach controls

For a Bot Coach you can edit, open its editor and find Proactive Nudges under Enabled Features. Set the consecutive-message limit, delay range, extra spacing, and optional prompt, then save. These are coach settings; they are not controls in the memory viewer. A coach you do not control may not offer editing.

Bot Coaches use an overnight quiet-hours window, normally midnight–06:00 in your account timezone, and also defer nudges during an enabled Sleep Mode window. Unlike AI Coach, the Bot Coach editor does not offer a per-coach quiet-hours picker or a separate messages-only nudge switch. Its permitted actions can run with a nudge, so review the coach's enabled features as well as the nudge prompt.

To stop normal nudges, switch off Proactive Nudges where you can edit that coach. Check Scheduled follow-ups too if you want to cancel planned check-ins. Disable Bot Coach is also available from an active coach's chat menu when Bot Hardcore Mode allows it, but disables the coach more broadly.

Stopping nudges does not cancel existing actions

Turning off nudges prevents future proactive messages once the setting takes effect. It does not cancel tasks, undo restrictions, or clear scheduled or queued device commands. Check those controls separately. See Shock Delivery & Queues for pending device actions.

Understand the timing

The delay range is not a guaranteed delivery time. Recent messages from you, a cooldown after coach messages, quiet hours, and extra spacing can push the next attempt later. Replying in the coach conversation resets the unanswered-message count; simply opening the app is not the same as replying.

Once the consecutive-message limit is reached, the coach waits for a reply instead of sending more nudges. Provider errors, usage limits, or an inactive coach session can also prevent a message. Quiet hours use the timezone stored on your account; without a valid stored timezone, that check cannot determine your local quiet window.

Scheduled follow-ups

Scheduled follow-ups lets AI Coach and Bot Coaches arrange a check-in for a particular time during an active lock session. It defaults to on when Proactive Nudges is on, including for existing coaches that have never set a follow-up preference. An explicitly saved off choice stays off. You can change it in AI Coach settings or the Bot Coach editor; follow-ups and normal nudges can also be enabled separately.

For example, ask: “Check in about this task in ten minutes.” The coach must use its scheduling action successfully; a promise in ordinary chat alone does not schedule anything. It can request a time from one minute to 30 days ahead and keep a private task note for the later check-in. At that time it reviews the current conversation and session before responding.

How follow-ups and normal nudges work together

Both use the same Shared autonomous message limit of 1–5 attempts before another reply (default 2). A pending follow-up reserves one slot. Normal nudges use only the remaining capacity, and a follow-up that is due takes priority. Reserving every slot for next week therefore pauses normal nudges until capacity is released; replying does not remove those reservations.

A new reply resets used attempts while keeping pending commitments. Failed or interrupted generation also uses an attempt. Cancelling a pending follow-up releases its reservation, but cancelling after generation starts does not refund an attempt. If you lower the limit, existing commitments remain and excess work waits for capacity rather than disappearing.

Follow-ups respect quiet hours and, for Bot Coaches, Sleep Mode. Active conversation, provider availability, and usage limits can also delay or prevent delivery. Automatic messages are spaced at least a minute apart; normal nudges retain their usual longer cooldown. An unstarted follow-up expires 24 hours after its requested time. This feature does not guarantee delivery at an exact instant.

Check or cancel a follow-up

Open Scheduled follow-ups in AI Coach settings or the active Bot Coach chat's settings to see pending times, deferred attempts, current allowance, and recent results. You can cancel pending items unless Hardcore Mode prevents it. An action already dispatched cannot be undone by cancelling its follow-up.

Ask the coach to change a pending follow-up, for example: “Move the task check-in to tomorrow at 14:00” or “Keep the time, but check whether I finished the report instead.” The coach can edit its time, its private task note, or both, without needing a spare slot. It can also cancel a follow-up when asked or when the commitment is no longer relevant. Check the recorded action result: a chat promise alone does not change the schedule. A running or finished follow-up cannot be edited. Disabling scheduled follow-ups, ending the originating session, or replacing its coach cancels pending work; re-enabling does not restore it.

The coach receives the current pending times and task notes, running state, and recent results in its prompt so it can manage existing commitments. Confirmations show a readable date and time in your browser’s locale and timezone, without internal schedule IDs. The private task note is omitted from chat action records and schedule status. Reasoning text and full debug payloads are hidden while scheduled follow-ups are enabled because they could repeat that note. The eventual message can still discuss the task itself.

Scheduled follow-ups use the same autonomous-action restrictions as normal nudges. AI Coach's Allow commands during autonomous nudges setting also applies to them: with it off, the follow-up can send a message but cannot run actions, schedule another follow-up, edit a follow-up, or cancel a schedule. Pending items can still be cancelled manually when permitted.

Read the schedule status

The schedule refreshes while it is open. Attempts used counts autonomous attempts since your last reply; reserved counts pending follow-ups. For example, with a limit of two, no attempts used, and one reserved follow-up, one ordinary nudge can still run. With two reservations, ordinary nudges wait for capacity.

Status	Meaning
A pending time	The coach accepted the follow-up. It has not started yet and reserves a slot.
Deferred; next check	The follow-up is waiting for conditions such as quiet hours, ongoing conversation, or available capacity. The next check is an opportunity to try again, not a guaranteed message time.
A follow-up is running	An attempt has started and uses allowance. Check the conversation for its result.
Completed	The follow-up message was saved. This does not mean you completed the task, or that every notification reached your device.
Cancelled	The follow-up was cancelled manually, by the coach, or because its settings or originating session changed.
Expired	It did not start within 24 hours of the requested time. Ask for a new check-in if it is still useful.
Failed	An attempt failed or was interrupted. It uses allowance and is not automatically repeated, because some actions may already have happened. Check the conversation and action results before requesting another attempt.

If a follow-up does not arrive

• Only a chat promise appears: look for a successful scheduling action and a pending time. Ask the coach to schedule it explicitly if neither appears.
• The appointment is deferred: check quiet hours, Bot Coach Sleep Mode, the shared allowance, and whether another conversation response is still running. Replying resets used attempts but does not remove pending reservations.
• The schedule is disabled or empty: check that scheduled follow-ups are enabled for the coach and that the original lock session is still active. Cancelled work does not return when you re-enable the setting.
• The panel cannot load or cancellation fails: reopen it to refresh the status. The item may already be running or no longer pending; do not assume a failed cancellation stopped it.

Accepted, unstarted schedules are retained through temporary service interruptions and checked again after recovery, provided they have not expired and the session is still eligible. An interrupted attempt is not automatically repeated. You do not need to send a message to trigger a valid pending follow-up.

Commands during nudges

A nudge can contain both a message and actions. AI Coach can execute actions when Allow commands during autonomous nudges is on. Bot Coach nudges can also execute actions, but its editor has no separate messages-only switch.

Actions blocked during autonomous nudges

For both coaches, the app blocks these actions during an autonomous nudge:

• Directly adding or removing lock time.
• Directly adding or removing cards.
• The dedicated action for adding time to an active pillory.

These restrictions apply even when the corresponding feature is enabled for ordinary replies. A custom nudge prompt does not override them.

Other actions remain available

The nudge restrictions are not a blanket ban on punishments or settings changes. The following action groups are not blocked by the nudge-specific filter:

Action group	Examples
Freeze and pillory	Freeze/unfreeze the lock; start pillory, remove pillory time, or end pillory.
Tasks	Assign saved or custom tasks, writing tasks, or quizzes; cancel an active task.
Devices	Shock or vibration commands, supported device modes, schedule changes, and stop commands.
Lock settings	Daily verification, casino games, card-pick intervals, cumulative chances, time/history visibility, rules, and supported editing permissions.
Hygienic unlock	Grant a temporary override or change settings when the separate permissions allow it.
Device Control and Programmable Lock	Supported phone restrictions and exposed program functions or variables.
Memory and images	Save remembered details or generate an image when available.

Each action still has its ordinary permission, session-state, device, consent, quota, and confirmation requirements where applicable. Availability does not guarantee that the coach will choose an action or that an attempted action will succeed. Check the recorded action result and current settings rather than relying on the message text alone.

Pillory has a specific distinction: starting pillory when it is already active can change its end time. Blocking the dedicated add-time action therefore does not prevent every possible pillory extension.

Can a nudge enable daily verification?

Yes. A coach can enable or disable the lock's daily verification requirement, including when it was previously off. That settings action is separate from the coach's Photo Verification permission, so disabling that permission does not prevent daily-verification settings changes.

Coaches have actions for specific built-in features, such as daily verification and individual casino games. This does not give them a general ability to install or toggle every extension.

What happens if I do not reply?

Nudges do not have a built-in rule that automatically applies a punishment after a missed reply or a response deadline. Once the configured consecutive-message limit is reached, the coach waits for another reply instead of automatically escalating.

Compared with ordinary replies, nudges have extra guidance discouraging repeated actions and treating old conversation as a fresh reason to act. Bot Coach nudges also discourage major punishment changes without a fresh, explicit reason. These instructions influence the model's choices; they are different from the actions the app blocks outright. Behavior can therefore vary with the model and prompt.

A prompt such as “punish me if I ignore you” does not create a reliable missed-reply enforcement rule or unlock blocked actions. Existing task deadlines, daily-verification requirements, restrictions, and scheduled device actions continue independently of whether you reply to a nudge. See Tasks & Task Queue for task-based requirements.

Bot nudge diagnostics and testing

In Bot Memory, expand Proactive nudge diagnostics. It shows the next scheduled check when available and recent outcomes. The scheduled check is an attempt time, not a promise that a message will be sent.

Result	What to check
Deferred: quiet hours or Sleep Mode	Wait until the applicable window ends.
Deferred: recent activity or bot cooldown	Recent conversation moved the next attempt.
Deferred: consecutive limit reached	Reply to the coach to reset the count.
Stopped	Check whether the coach/session is still active and Proactive Nudges is enabled.
Failed	Read the provider/error details, then check the model, provider availability, key, and usage allowance.
No scheduled nudge / schedule unavailable	These are different states: one has no scheduled entry, while the other could not retrieve schedule information. Neither alone proves the provider is broken.

Run nudge test bypasses the normal timing checks for one attempt. It still requires an active Bot Coach with nudges enabled and a working provider. Tests are limited to three per hour per lock.

A nudge test sends a real message

The test is not a preview. A successful attempt appears in the conversation, uses the configured provider, and can run permitted coach actions just like a normal nudge. Review permissions before testing, and check the new diagnostic result instead of repeatedly pressing the button.

If memory seems wrong, inspect Stored Memory, correct the specific fact in chat, and verify the saved result. If a Bot memory refresh appears stuck, wait for the background update and check provider limits before retrying. If nudges stopped, first check the active session, enabled feature, consecutive limit, and quiet/Sleep Mode windows before changing provider settings.