What’s new with Murph

Murph now keeps delayed wearable imports moving alongside routine background work, helping connected activity and health data reach follow-up work without repeated delay.

Recovery stays bounded so retained imports can catch up without holding up other background work. Interrupted progress checkpoints now converge on the next pass instead of causing repeated retries, and Murph does not mark data processed until the import succeeds or reaches a confirmed failure.

Murph now answers ordinary onboarding and setup questions directly, reserving early progress notes for work with several substantive checks or a longer wait.

Routine context reads and one or two quick calls stay in the main reply, while longer research, document inspection, and other substantial work can still receive one useful update.

You can now delete your Murph account from Settings without first creating a passkey or secure approval wallet.

Deletion still requires a signed-in session and the exact confirmation phrase, and remains available if secure approval is unavailable while vault export stays protected. Existing provider cleanup, retry guidance, and post-deletion sign-out remain unchanged.

Murph now checks relevant saved details before asking you to repeat them and keeps working through ordinary website controls during appointment check-in.

Murph still pauses for a real consent choice, private credential, CAPTCHA, or uncertain submission, and only saves sensitive details when a proper canonical owner exists.

When you explicitly ask Murph in a supported group chat to consult your own private Murph, it now follows whether you want the answer in the group or privately and asks when that is unclear.

The request stays tied to you and that group. Choosing group delivery keeps the existing advance notice; choosing private delivery does not post that notice. Existing permission and privacy checks remain, and the group never gets direct access to your private Murph or personal vault.

When you tell Murph in a private chat that the action for a current reminder is already done, that occurrence now stays quiet instead of repeating the cue.

Only the matching occurrence is skipped; future reminders remain active, and unclear or older completion messages do not suppress a cue.

When Murph teaches a movement in a scheduled Messages cue, it now includes a reviewed exercise picture when one is available and keeps internal catalog labels out of the conversation.

If a picture is missing and you ask to see it, Murph uses the reviewed catalog image when available and can create an instructional image when the catalog has none. Setup-only and non-instructional turns remain text-first.

New recurring reminders now begin with the reminder you asked for, without immediately asking whether to keep, change, or pause it.

Murph asks about cadence only when a confirmed earlier cue has no relevant reply. Prescribed treatment, clinical monitoring, and safety-critical reminders keep their existing protections.

Scheduled group health updates now keep each participant's name with their shared values, and current-value questions refresh those labels before answering.

Labels come only from consented shared rows. A new snapshot is not used to remap different earlier figures, identify a sender, or expand what the group can access.

When you ask Murph to bring a private analysis into a group, the shared update can use your group name and stays available for the room's follow-up questions.

Murph uses only the group-safe name and private context you chose to share. If the consultation cannot finish, its response stays tied to the earlier question instead of arriving as an unexplained message.

If you send a follow-up while Murph is preparing a response card, Murph can now attach a fresh card to the latest reply instead of falling back to text.

A newer message still clears any card chosen for the earlier request, and the fresh card must fully answer the current request.

Murph now rebuilds its private workspace index without redundant audit copies and unused lookup structures, reducing backup bulk while keeping audits and exports complete.

Canonical health and audit records are unchanged. Older closed event history now compresses automatically during interruptible idle checkpoint maintenance, while new messages still preempt cleanup.

When Murph's hosted assistant wakes after idle time, it now skips unnecessary setup work and retries a stalled startup check sooner before working on your message.

Older saved state still receives full compatibility checks. Overall startup and health checks remain in place, and unsupported state or true startup failures still stop safely.

If you choose to connect a wearable later, Murph now continues with the next useful onboarding question instead of pausing the whole conversation.

The connection remains optional, and Murph does not claim that wearable data is connected or syncing until the connection is actually visible.

If a brief account update collision happens as Murph reaches its hosted usage limit, the one-time explanation now retries instead of disappearing.

Only the pre-delivery claim retries, and only once; message delivery itself is not repeated.

When you approve a prepared file in a private Murph chat, Murph now sends it without waiting for another message.

The approved attachment stays bound to its original private conversation, and completed temporary copies are cleaned up without touching saved vault files.

The topic checklist is now the largest text on the screen, the repeated setup copy is gone, and the live transcript no longer cuts a line in half.

Reopening the interview no longer repeats a detail you already answered in the same session, a spoken range such as "twelve to fourteen hours" is saved as a number, an imperial measurement is converted before it is stored, and temperatures read in the unit system your device reports. Letter grades now follow the A to F scale.

Short workout comparisons now fit into one shared-header table instead of repeating every field vertically.

Wider content still uses the full-width layout, and the complete text version remains available when an image cannot be delivered.

Murph now gives each background connected-data pass up to two minutes to finish unusually slow sync jobs.

Foreground replies still take priority. Longer-running syncs stay recoverable, and Murph can now pinpoint slow background phases without putting account identifiers or health values in operational logs.

Overview and History now keep readable data during refresh problems and show a clear unavailable state when dashboard data is not yet ready.

Background checks stay bounded and the dashboard updates when a later snapshot becomes available, without asking you to manage repeated attempts.

Links can now open a specific connection on Murph's Connect Devices page.

The requested source opens in a focused dialog; sign-in and authorization still require the member's explicit action.

Long connected-health histories no longer hold up newer sync work.

Murph saves each completed batch and resumes at the exact next day; timed-out background requests also yield to newer updates, while between-sync corrections still receive a full safety recheck.

Food search now bounds catalog work before ranking results, reducing timeouts for broad terms.

Exact names, brand-only terms, generic foods, and typo recovery keep their existing paths while ranking remains intentionally bounded across the catalog.

When Murph joins an existing group chat, that group can now appear in its owner's private Murph without a separate join-link step.

Only the group owner is added automatically. Other participants still join explicitly, and Murph does not guess between multiple unnamed groups.

When you ask private Murph to share an update with a group, the group message now keeps your actions and claims attributed to a member instead of presenting them as Murph's own.

Murph uses the sender's saved preferred name when available; otherwise it uses a neutral member reference rather than guessing.

Home now reduces overlapping data work so dashboard details are less likely to fall back during busy periods.

Independent sections still recover separately if one optional detail is unavailable.

Murph now waits five days before checking on a quiet supported wearable, and you can ask Murph to wait longer or stop checking.

Missing data by itself is no longer described as a disconnected account; reconnect guidance is reserved for an authorization problem.

In an existing iMessage conversation, Murph now prioritizes your new message over background connected-data work.

If background work is already saving, that save finishes safely first; anything left resumes after your reply.

When you reply to an earlier Murph message in a direct conversation, Murph now uses that exact message to understand what you mean.

The selected message is treated as the reply target without changing ordinary conversation context or tapback handling.

Scheduled cards now arrive when the finished card is the complete response.

This covers private nutrition and workout updates plus existing authenticated group cards. Complete text fallbacks also keep moving when a rich card cannot fit.

When a group requests many health permissions, Murph now summarizes them in seven clear categories instead of sending a long item-by-item list.

The reaction still applies the exact permission set. Narrow offers remain specific, and the review link lets each person inspect or customize every permission before sharing.

When an automatic meal photo is too unclear to estimate, Murph now asks one focused question instead of leaving daily nutrition stuck.

The retained image still follows its privacy cleanup schedule. A reply updates the existing meal instead of creating a duplicate, then Murph refreshes the nutrition totals and eligible card.

Videos sent to Murph can support the current request without being carried into hosted workspace snapshots.

Raw video bytes become eligible for removal at the next safe idle cleanup after active work no longer needs them. Attachment details and transcripts can remain so the conversation still makes sense, while videos explicitly saved as durable evidence keep their existing lifecycle.

Answers saved during the Environment interview now reach your report even when Murph is already handling another request.

Your save stays durable while current conversation work finishes, then the next mailbox pass applies it in order and refreshes the report.

Completed file deliveries no longer add avoidable weight to later conversation restarts.

Files still needed for delivery remain protected while completed temporary copies can be cleared safely.

When a group-to-private handoff needs one more detail, Murph can now ask and continue after the participant replies.

Fresh private requests and follow-up answers use their existing separate paths; queued work is still not described as sent.

When several connected apps report workouts, Murph now keeps detailed workout sync moving even if one source does not provide that data.

Existing workout facts from sources without that detail remain intact, while sources that support it keep adding their available workout signals.

Murph can now complete more nutrition cards with compatible saved goals and recover a captured meal when its first update fails.

Saved goal meaning stays unchanged, and automatic captures retain a photo-supported observation when a trustworthy numeric estimate is not available.

Murph now confirms saved recurring reminder changes without making ordinary scheduling work sound like a failed update.

Active recurring reminders stay active while the next occurrence is finalized automatically; genuine timing problems remain explicit.

Setting a reminder no longer fails because an older reminder has the same name.

Each new reminder now gets its own stable identity, while existing reminders remain editable by their exact identity.

Messages that arrive while Murph is restarting after a service update now move into recovery without an avoidable retry delay.

Recovery still waits for the current conversation state to be saved safely before a replacement begins, so the faster handoff does not trade away context or duplicate replies.