What’s new with Murph

Murph can now retain bounded cadence, power, heart-rate, split, distance, duration, and swimming features from supported workouts.

Workout context can now include compact heart-rate, cadence, power in watts, speed in meters per second, and split details for each connected workout. Murph does not retain coordinates, raw points, complete curves, or provider response arrays.

Murph can now retain timed insulin and carbohydrate records alongside a more useful daily glucose summary from connected sources.

Glucose stays aggregate-only: Murph keeps bounded variability summaries, not the full provider timeline or raw samples.

Murph can now retain connected bone, muscle, visceral-fat, body-water, weight, body-fat, BMI, lean-mass, and waist measurements.

Murph accepts only documented units and valid dated records, keeps normal correction history, and does not retain provider timelines or raw sample arrays.

Workout explanations now keep complete movement sequences within a bounded image set so the whole response is less likely to fail in Messages.

Murph prioritizes the movements that benefit most from setup, transition, and endpoint frames. If a direct-chat image response fails, it remains an image task instead of being replaced with text.

Exercise routines, training plans, schedules, and other structured answers now use the available card or movement media instead of a long plain-text layout.

Telegram uses Rich Messages when a structured layout makes the complete answer easier to scan, including when no standard Murph card fits. Exercise routine cards include available catalog images by default, unless you ask for a routine without them. Existing nutrition, summary, and exercise cards keep their familiar layouts. Messages pairs its available card or ordered movement media with the concise instructions needed to complete the plan safely. Murph uses concise text alone when a richer layout would not help.

Murph can now retain bounded within-day oxygen and stress patterns when connected-source readings are close enough to support them.

It keeps sample-counted oxygen burden plus run, variation, and daypart summaries only when nearby readings support them; isolated readings stay out of continuity claims, and raw samples and full timeseries remain excluded.

Murph now keeps new messages in front of background wearable catch-up work, reducing avoidable reply delays while a connected device syncs.

Wearable catch-up remains retryable and resumes automatically, while normal connection checks keep their regular schedule.

After finishing account setup in the companion app with a verified phone, Murph can now start the conversation with the standard first iMessage.

When a healthy line is available, Murph sends the usual welcome and may keep unfinished setup moving with a low-pressure check-in on the next few days. If a line cannot safely start the conversation, setup still completes, and texting an active Murph number starts it from your side.

Murph can now use sleep latency, separate activity-intensity minutes, activity-specific heart-rate fields, and bounded profile and cycle facts from connected health sources.

Activity intensity keeps the source's minute values, and activity heart-rate averages remain distinct from sleep averages. Newer complete profile and cycle snapshots can replace changed source facts and remove omitted ones, while member corrections remain in control. Invalid dates, predictions, unfamiliar values, and legacy length fields remain bounded supporting evidence.

Personal Patterns now shows distinct illustrations for activities and tags, while phone layouts keep three next-day outcomes visible without sideways scrolling.

Unrecognized activities and custom tags use neutral illustrations. Pattern calculations and evidence thresholds are unchanged.

Murph can send approved Gmail and Outlook messages and add primary-calendar events through the selected connected account again.

Email still requires a current private request. Primary-calendar actions can also follow an exact direct automation the member saved. Every action uses a selected account, and Murph does not repeat an uncertain provider result.

Connected health sync now requests every measurement Murph supports, including steps, distance, active calories, heart rate, and weight.

Murph keeps dense signals compact: steps and distance are daily totals, active calories and heart rate are hourly summaries, and weight remains an individual reading with longer history.

The modern Dexcom card now says when production connections are coming soon instead of opening an unsupported setup flow.

Existing connected accounts keep their management controls, and the separate Dexcom G6 and older connection remains available.

When a connected health source refreshes personal data, consented group views now pick up the newer shared snapshot on their next read.

Fresh messages keep priority. Murph checkpoints first and keeps personal sync and group sharing decoupled, without widening any sharing permission.

Local Murph setup now focuses on Telegram and no longer asks you to configure a separate email inbox.

Hosted email conversations, connected-app email actions, and Murph account or service emails are unchanged.

When you ask for a one-time reminder in a named timezone, Murph now resolves that local date and time before saving it.

Words like tonight and tomorrow stay tied to when you asked, even if processing crosses midnight. Murph confirms the stored schedule and next deliverable occurrence before reporting or changing it, and clarifies nonexistent or ambiguous clock times instead of guessing.

After Murph processes an environment recording, the Environment page now requests a fresh private report and keeps showing progress until that newer report arrives.

If processing or the private report takes longer, the page offers Check again without asking for another recording.

Murph now syncs dated Oura sauna tags so repeated sauna entries can become Personal Patterns factors.

Wearable tags stay neutral notes. Personal Patterns currently treats only Oura's exact sauna tag as an action; symptoms, context, custom tags, and tags from other sources stay neutral. Existing Oura tag history receives one bounded refresh under the corrected meaning. Murph does not retain or analyze the provider's free-text note value.

Personal Patterns can now compare repeated wearable activity with selected next-day recovery metrics when enough evidence exists.

The existing matching thresholds stay unchanged, and unclear or insufficient evidence still remains labeled as such.

When you ask Murph in a group to continue privately, the reply now stays in your existing personal conversation so you can follow up without starting over.

Murph still verifies the current sender, same messaging channel, direct route, expiry, and exact reviewed answer before delivery, and never posts the private answer to the group.

When a group asks someone's personal Murph for help, Murph follows the requester's ordinary wording to answer here or privately.

If the audience is genuinely unclear, Murph asks a normal follow-up instead of requiring special phrasing. Before opening personal context, Murph fixes whether the reviewed answer returns to the room or the sender's existing direct chat; when it will return to the room, Murph tells the room first. Every request stays bound to the exact sender and message, retries cannot complete it both ways, it expires within ten minutes, and it grants no future access.

When you approve an exact health share in a Murph group, Murph now starts preparing it right away and tells the group while the shared view is still getting ready.

Pending shares stay unranked and do not trigger another permission request or a guess about device status. Reapproving a recent-data share starts a fresh preparation instead of briefly reusing an older view, and preparation retries safely if the first attempt fails. Sleep-sharing prompts also state that sleep timing and duration cover the last seven days, with the same exact permission and privacy boundaries.

Connected wearables can now bring up to six months of supported low-frequency history, including AFib burden, VO2 max, temperature, breathing disturbance, recovery, hydration, caffeine, and mindfulness.

Dense streams stay bounded to recent data, and the longer import runs in the existing background sync only for data the connected source says it provides.

Settings export now downloads the newest retained dashboard data Murph has instead of blocking while newer Oura or other health changes are still processing.

The export explains that recent changes may be absent and requests a background refresh only while health-data consent remains granted. A missing retained copy stays retryable without consuming the failed attempt's one-time challenge.

Murph's web deletion page now explains how to remove an account even when the Android app is no longer installed.

The page keeps deletion controls available without an active subscription or renewed health-data consent, and names what the request covers, category-specific retention targets, external-delivery limits, longer-lived support, security, billing and tax records, and where to request help or partial deletion.

Static nutrition, workout, table, and challenge previews now match their iPhone layouts more closely and carry Murph's mark instead of a blank corner.

Cards keep their full rows and progress states, align each title beside Murph's mark in one shared header, and keep nutrition status in the safe text fallback instead of repeating it beneath each value.

Murph can now recognize and use more body-composition, respiratory, metabolic, cardiac, safety, accessibility, and workout signals from connected health data.

Murph checks the available normalized facts instead of assuming a source supplied them. ECG waveforms and workout streams remain compact summaries, and missing coverage stays unknown rather than becoming zero.

iMessage conversation previews now say whether Murph sent daily nutrition, a workout, a general summary, or challenge standings.

The preview label stays free of card values in the conversation list, on the lock screen, and in notifications, while keeping a short way to ask Murph for the card in text.

On phone-size screens, the public homepage now gives its iPhone conversation preview a narrower, more natural shape while keeping the message and signup action easy to read.

Tablet and desktop proportions, conversation behavior, and the signup path are unchanged.

Connected health syncs now preserve more daily heart-rate, activity-intensity, and sleep-latency details as separate structured measurements.

Murph keeps low, medium, and high activity minutes independent, preserves the existing total, and converts provider sleep latency into canonical minutes.

Connected health syncs now preserve bone mass, muscle mass, body water, and visceral fat readings as separate structured measurements.

Murph keeps the upstream provider and units attached to each fact, omits missing values, and leaves profile height with its existing owner.

Connected health syncs now preserve more dated menstrual observations and the profile gender field as structured personal context.

Murph keeps provider predictions excluded, records categorical observations without treating them as diagnoses, and keeps gender separate from biological sex.

If a printer rejects a physical note, Murph now explains whether to check the address, remake the artwork, or wait for a Murph-side fix.

Nothing is claimed as sent, uncertain submissions are not retried, and printer details stay out of the conversation.

Eligible private meal logs and verified live-workout mutations now return their refreshed card in the same response, including ordinary meals and incremental set updates.

Murph uses the card as the complete reply when the owning safety, authority, evidence, route, and bounded-card checks pass; otherwise it falls back to truthful concise text.

Messages can offer Murph from the App Store when the app is missing, while opening a shared card on the web shows the same short handoff.

The page keeps the card details opaque and leaves the shared link unchanged. Ordinary homepage links stay unaffected, while Messages continues to own the interactive card and its install prompt.

Settings now presents Starter usage and unavailable models more concisely while keeping relevant plan and upgrade choices visible.

A never-paid Family setup no longer traps a later invite, and Settings now clearly separates starting a plan you pay for from joining with someone else's invite.

Murph removes only an inert owner-only draft as part of successful invite acceptance. If a Checkout is still live, Settings explains the conflict and can expire and abandon that exact unpaid setup before the member retries the invite.

Referral reward celebrations now stay attached to their intended direct conversation, and an interrupted one can safely resume without leaving later work waiting behind it.

Murph resumes the already-pending celebration itself and rechecks the original direct conversation before any new response work. It never switches to a newer route; if the original conversation is no longer authorized, that celebration ends without sending so later notifications can continue.

When Murph is recovering older blood-pressure readings, a temporary change in source access no longer lets an unfinished history import look complete.

Murph keeps the same history window available to retry after access returns and marks it complete only after every remaining day finishes under the currently connected source.

Pending wearable updates can now finish their sync work even while scheduled assistant automations are paused.

Fresh messages still take priority. Murph checkpoints recovered health-data changes before marking them handled, retries safely when a source remains pending, and does not start an AI response just to finish the sync.

When you schedule a recurring reminder in a named timezone, Murph now preserves that local time through saving and later edits.

The confirmation comes from the saved schedule and scheduler's next deliverable occurrence. If timing cannot be verified, or an old one-time reminder can no longer fire, Murph says so and offers a bounded recovery instead of inventing a time.

When Murph compares a message with connected health activity, it now sees the local clock beside the exact UTC instant instead of having to relabel a raw timestamp.

That makes timing explanations more reliable across daylight-saving changes. Wearable imports also separate provider-send, Murph-receipt, and successful-import timing when those clocks are available, without putting health values in operational logs.

Ordinary voice memos now use the voice already selected for your Murph instead of switching voices on their own.

A different named voice is used only when you explicitly ask to test it or request that voice for one memo. Saving a named voice and asking to hear it immediately still works as a one-time preview.

Across Luna, Terra, and Sol on OpenAI, Murph now reviews one available initial photo with more detail.

Multiple available initial photos, follow-up photos added while Murph is working, Venice, and custom inference stay at high detail to keep visual input cost and compatibility bounded.

Workout detail and nutrition goal direction now stay inside their cards instead of repeating a long summary beneath the static image.

Fallback cards have no duplicate badge or corner mask. Nutrition goals use short in-card direction labels without repeating target amounts, older sent workout cards remain readable, and generic tables keep their optional subtitle.

When Murph uses managed OpenAI, its built-in web search can reach current information again instead of stopping with a forbidden-request error.

Search still runs through Murph's existing protected managed OpenAI provider connection and returns in the same conversation. Other provider choices keep their current search behavior.

When a future care appointment is confirmed in a private conversation, Murph now creates one useful reminder by default unless you opt out.

Morning appointments use the prior evening, later appointments use the same morning, and Murph keeps the same reminder up to date when an appointment is rescheduled or canceled.

Authorized scheduled nutrition cards now use the member's current private Messages conversation, so the native card can arrive without a new inbound message.

The same live destination check covers ordinary replies, cards, voice memos, and reactions. If the route changes before delivery, Murph stops instead of sending to a stale conversation or exposing private text through recovery.

Completed exercises now keep a clear checkmark in Messages workout cards, including their static previews.

The status mark is part of the card image itself, so it stays visible anywhere the static preview is shown.

Health data use now keeps source and consent actions aligned with the other Settings controls on larger screens while staying compact on phones.

Manage sources and Withdraw consent keep their existing destinations, confirmation, status, and recovery behavior.

The public homepage now uses compact avatar images while continuing to prepare secure sign-in automatically in the background.

Cold sign-in still opens immediately, announces its loading state, keeps keyboard focus when the form arrives, and retries a temporary loading failure. Text contrast is also clearer across the updated sections.

The private Environment report now opens with a clear preparing state and a report-shaped preview instead of a mostly empty page.

The preview mirrors the printable report and gives immediate feedback while the existing private Browser Vault opens. The finished report, empty state, and error recovery continue unchanged.

Murph now checks the current shared sleep record before answering, counts reported Deep and REM sleep as soon as those values are shared, and includes explicit manual corrections.

The latest manual correction for a sleep date wins and is labeled Manual instead of a connected source. Reconnected sources no longer combine an old disconnected status with a newer sync time. Future-dated entries stay excluded, and missing data remains unverified.

Direct connections now explain the handoff before authorization, Apple Health relay sources use recognizable service icons, and unsupported sources get verified export guidance.

Relay cards still make Apple Health ownership clear. Manual exports are described as snapshots rather than live sync, private files move to your private Murph, and Murph does not invent an export menu it cannot verify.

Murph now routes weight loss, weight gain, cutting, bulking, recomposition, and maintenance into one evidence-backed body-composition approach.

The guidance keeps the roles of nutrition and training clear, compares trends in consistent units, minimizes tracking burden, and uses separate safety handling when pregnancy, postpartum recovery, breastfeeding, or under-fueling changes the answer.

Murph now understands the native reply relationship between group messages and gives ordinary conversations a longer beat before answering.

A reply edge never invents the target's words or identity. Human-owned beats can finish silently, urgent coordination skips avoidable delay, and a burst still produces at most one terminal action for the room's current moment.

Funding a group no longer produces a song or other public creative response unless the payer opens the optional personalization and chooses one.

Message, poem, and 15-second song are explicit formats. Usage credit is granted independently of creative success, and missing or older customization state never counts as song consent.

Private nutrition and compact-table cards now keep their full structure when a long conversation is condensed and still use Murph's verified web address.

The result returns as one card in the same conversation with an accurate static or text fallback. Empty fallbacks stay silent, and invalid card data is rejected.

Static nutrition cards now use Messages' own app icon and rounded frame, with only the date and meal count beneath the card unless totals are partial.

Calories, nutrient totals, and goal status stay inside the card without a second Murph badge or a long repeat below it. Provider chrome keeps only a short partial-data warning when needed.

For an established private chat, an authenticated typing hint can help Murph get ready while you compose the message.

Typing is never treated as a message, permission, or delivery claim. Unknown or ineligible hints quietly fall back to the normal path, and the later accepted message still owns the reply.

Dynamic recurring automations can now use a small recent-output history to avoid substantially repeating the same quote, fact, prompt, suggestion, or recommendation.

The history stays tied to the current version of the automation. Exact-text reminders remain exact, and old output is treated as history rather than a new instruction.

Murph's public footer now includes a direct link to the official iOS App Store listing alongside the existing product links.

The link opens in a new tab with the existing external-link safeguards and keeps the current footer hierarchy intact on desktop and mobile.

Available on iOS

Murph

Open the App Store listing

When Murph restarts while saving its latest work, the new instance now waits for that exact handoff instead of restoring an older conversation.

The handoff remains automatic and limited in time. A faster save preserves the same protection for restored and newly arrived messages without adding routine startup delay.

Pausing Murph no longer blocks already-authorized cleanup of expired private inbox media.

Ordinary replies stay paused. Only already-approved retention cleanup can briefly resume, remove expired media, and save the cleaned state.

Proactive follow-ups and completed phone-call results now do less blocking background work, leaving current replies more room to run during busy moments.

Current replies still keep priority, while card and call destinations, duplicate protection, and recovery stay unchanged.

When a product problem has useful reproduction evidence, the internal feedback path can store a bounded model-written summary with concise steps and environment context instead of attaching the raw conversation or service response.

Ordinary feedback remains silent and best-effort. Recognizable shaped identifiers and secrets receive deterministic scrubbing, but the path does not claim that every private meaning can be detected or removed.