Changelog
The latest seven days of features and improvements, with the full archive one step away.
Environment turns what Murph knows about your home into a private report and one-memo walkthrough. Group funding and access, link previews, reminders, usage, Venice replies, and delegated work all get clearer or more reliable paths.
New features
The new Environment page turns saved facts about sleep, air, light, recovery, and workspace into a private report with category notes, live local conditions, and an A–E audit once enough scoreable conditions are known.
Missing or skipped facts never lower the grade, and optional equipment never counts against it. Each result explains what is known, what needs attention, and the next useful check.
Walk Murph through five short topics in one recording. Murph privately extracts clear facts about your home and workspace, saves them to your vault, and refreshes the open report when processing finishes.
A partial report asks only about useful missing details, while a complete report switches to a free-form update. Precise addresses are rejected, ambiguous details stay unknown, and the transcript is not added to conversation history.
In a supported existing chat, a reply with text or media followed by one HTTPS link can send the content first and the link as a native preview.
A link-only reply can use one native link when it does not need a reply anchor. If the preview is definitively rejected, Murph sends the exact link as text, and retries preserve content that was already accepted.
An active Family owner can choose Add usage from their own AI-usage row and select $5, $10, or $25 through the existing Family checkout.
A fulfilled purchase starts a fresh display window at 0% used, and only later usage moves the meter. Each payment return stays with its exact owner or member surface instead of creating competing confirmations.
A group can ask Murph to create or extend access through one action. Supported iMessage groups keep the native reaction path, while SMS, Telegram, scheduled turns, and standalone requests get the existing first-party link.
Murph chooses the presentation from the trusted route, includes one usable access surface, and returns a truthful unavailable result when the route cannot be proven.
Anyone in an active hosted group can open its first-party funding page at any capacity. Unsponsored groups open monthly sponsorship, while other participants in sponsored groups can add a one-time contribution.
An explicit request to fund, sponsor, contribute, or get the funding link no longer detours into referrals or claims that a healthy group needs funding. The active payer can manage sponsorship on the same page, and every payment still requires explicit confirmation and Stripe reconciliation.
Improvements
A one-time reminder created near its scheduled minute now keeps a durable wake with the saved schedule, so ending or restarting the creating session cannot strand an active reminder.
The current reply confirms the saved reminder without waiting on the best-effort wake signal. The scheduled message keeps its existing route, timing, and delivery checks.
Members who choose Venice can now receive Murph's normal tool-enabled reply instead of seeing typing end when the provider rejects the request shape.
The model choice, tools, conversation, and delivery path stay the same. Murph translates only the exact supported tool envelope and rejects malformed or conflicting metadata before provider entry.
Eligible group members now see Core in Settings, trial continuation, billing recovery, usage-limit messages, Clubs pricing, and private plan answers.
Only the name changed. Core remains $3.50 per month with the same included usage, eligibility, billing transitions, assistant capability, and group continuity.
The usage meter now answers how much capacity remains, while waiting, active, final-checking, and reward-pending referrals stay in the current list until they are complete.
A quiet Details row holds each referral's requirements and selection date. Only completed referrals and usage purchases move into History.
Murph now judges a proposed group dare by the concrete act, so an ordinary timed stake or playful consequence can stay intact when it is safe and consensual.
The wording alone does not trigger a veto. Real hazards, coercion, impairment, contraindications, or pressure through distress still get the narrow boundary they require.
Hosted experiment progress cards now keep rendering when optional biomarker-direction context is unavailable, using neutral mover sentiment instead of failing the whole request.
The card visibly says when direction context was unavailable, and the same accessible description travels with the private image in supported iMessage and Telegram delivery.
When you ask Murph to handle, choose, decide, or figure something out, it now completes everything useful that is independent of a blocker before asking for more input.
If a texting reply still needs a decision-changing fact, Murph asks one highest-value question at the end. Delegation still cannot create new permission to spend, contact, publish, schedule, or take another external action.
Capped group sponsorship joins the ways to get more usage, experiments can end in more than a number, and group conversations gain sharper voice, GIF, name, and app-download context. Sign-in, checkout, wearable setup, and cold replies all get a cleaner way through.
New features
A current participant can authorize up to $5, $10, or $20 per month for a group. Murph starts with one $5 purchase and makes another $5 purchase only when the group needs capacity and the private maximum still allows it.
The sponsor can privately change, pause, resume, recover, or cancel the authorization. Unused purchased credit stays with the group across sponsorship periods, automatic refills stay silent, and a separate one-time contribution remains available.
Ask Murph how to get more usage and it now checks the available plan, top-up, group funding, and ways to earn usage together. Settings also shows recent credits and each current or completed mission.
Choosing one or both missions remains explicit. Each tracks and can be cancelled independently, and a mission past its action deadline reads as final activity being checked rather than still actionable.
Murph can now save custom experiments with numeric, photo, document, or structured-review outcomes when the intervention, capture route, and review timing are clear.
Numeric outcomes use their declared comparison. Qualitative evidence gets an honest before-and-after review receipt instead of a made-up numeric effect, and incomplete evidence stays recoverable.
Eligible personal members can choose OpenAI or Venice beneath the Luna, Terra, and Sol model choices in Settings, then save both choices together.
The compact setting always says which provider is saved and when a draft will take effect. The choice changes future core replies only; specialized tools keep their own managed providers.
Murph can now help with assignments, essays, exam questions, drafts, and educational code in the current private or group conversation.
The boundary follows the purpose of the request: schoolwork is in scope even when the subject is professional, while production code, client deliverables, and operational work stay out.
When a group clearly keeps a playful jab going, Murph may answer with one short, self-deprecating voice memo and no duplicate text.
A stop request, a correction, a human-owned beat, or ambiguous hostility still means silence or a plain reply. Saved Humor 0 disables the unprompted sarcastic memo.
Animated GIFs sent through a supported iMessage conversation now become a compact left-to-right filmstrip that Murph can inspect in the ordinary reply.
Frame selection follows playback time so held reactions and punchlines survive. Malformed or oversized animations keep the existing unavailable-attachment path and are not stored.
Improvements
Personal, Family, and group funding now present dollar-denominated cost-weighted usage credit without converting it into an approximate number of messages.
Actual model and tool costs vary, so the credit amount is the durable financial truth. Group participants see only whether Murph is sponsored, while exact charges, pending purchases, and monthly maximums stay private to the sponsor.
Ask Murph how to download the iPhone app in a private or group conversation and it now shares the canonical App Store listing in the same reply.
The public download link can be shared with the room, while personal sign-in, Apple Health authorization, and wearable setup still stay private or in the app.
In a private hosted chat, Murph now uses the current time with your saved timezone before suggesting an immediate meal, workout, caffeine choice, or bedtime step.
Group Murph can reason about the room's current clock without treating the room timezone as any participant's personal local time.
Several group messages arriving together now become one natural Murph turn and one reply, with safe participant names used naturally when they are available.
Names remain presentation help, never identity or permission. Explicit alternatives stay visible, and participant-specific actions still resolve from the exact source message.
Opening Signup or Log in from a cold homepage now shows the ordinary phone, Telegram, and email form immediately instead of replacing it with an inert loading state.
The chosen method becomes busy at once, waits for the shared auth runtime when needed, and keeps the same action in place through account setup and consent.
A hosted wearable connection now returns to the same signed-in browser that started it and waits for an explicit Finish connection before attaching the account.
Adding or retrying one source leaves established sources active. A pending or disconnected source stays inert until the owner-bound confirmation succeeds.
Starting or retrying standard or Pulse Checkout now returns to the same open Stripe session instead of creating a competing payment path.
Only the first valid completion can become the subscription. Conflicting, stale, or ambiguous ownership keeps the existing recovery guidance rather than starting another charge.
Scheduled group newsletters now use the seven completed local days before the run, regardless of weekday, and continue delivery from one accepted edition.
A restart or later receipt repair no longer recomposes the newsletter. If no usable completed-day stats exist, the edition says only that.
The one-time contribution dialog keeps optional context under Add a note and turns verified success into a clear receipt with Open Messages.
Amount choice, payment verification, recovery, and group-credit delivery keep their existing owners. The receipt stays honest that Messages opens at the app level.
When a supported message wakes a stopped hosted Murph, it can begin warming up while the message finishes arriving, shortening the cold path before the same reply starts.
The reply, conversation history, model, and delivery path stay the same. If the early start cannot be reused safely, Murph falls back to the ordinary startup path.
In a playful group turn, Murph can do one brief public lookup when it needs the premise or vocabulary for a specific reference-native reply.
Known references add no lookup. If the reference stays unresolved, Murph replies plainly instead of bluffing, and it never turns the joke into a research summary.
iMessage edits become real corrections, dense reminders become a natural reply loop, group challenges gain team and multi-metric scorecards, Telegram can hand off the right iMessage line, and onboarding, clubs, group context, images, experiments, and the public site all get clearer.
New features
Around three weeks after answered onboarding, Murph can send one low-pressure question about what feels worth improving, understanding, or handling now, or stay quiet when the context is not useful enough.
It is a one-time choice point, not a recurring nudge. Existing eligible members get one future catch-up rather than an immediate late message, and nothing changes until they reply.
One-time follow-up
Murph can now run individual, team, or whole-group challenges with up to five additive scoring components, including weighted combinations such as steps, logged protein, and qualifying workouts.
Rules and team membership freeze before scoring. Missing data stays visibly partial instead of becoming zero, and cumulative challenges can roll forward without rewriting earlier results.
The Clubs page explains collective, team, and head-to-head challenges, shows how iMessage and supported wearables keep standings current, and makes it easy to email Murph to start.
Organizers can preview setup, scoring, private member support, and launch details without a spreadsheet. The page stays public, and the email address remains visible if a mail app cannot open.
Start a club challenge
When the current group title matters, Murph can now read it on demand and carry the exact name into a new-group setup instead of guessing.
Missing, synthetic, or unavailable titles stay unnamed, and the title is treated as display text rather than an instruction.
In a private Telegram chat, ask Murph for its iMessage number and it can return your existing assigned line or reserve one healthy line for the exact verified phone on your account.
A success names the masked phone that should send the first iMessage. Email-only, mismatched-phone, or unavailable-capacity cases keep Telegram working and point to Settings when account proof needs attention.
Improvements
When you ask for several same-purpose reminders in one day, Murph can offer one finite check-in loop that asks naturally about the previous action while cueing the current one.
It carries forward at most one unresolved occurrence, never builds reminder debt, and goes quiet after one unanswered combined check-in until you re-engage.
When you edit a message, Murph can use the corrected wording in the active or next turn instead of silently keeping the first version.
If Murph already answered, it follows up only when the edit materially changes the answer or action. Typo-only edits stay quiet.
iMessage
When someone is added to or removed from a supported iMessage group, Murph can carry that bounded context into the next ordinary message instead of losing the change or announcing it on its own.
Duplicate events do not restage it, and optional contact labels remain advisory rather than identity.
After accepting a direct image request, Murph now says it is making the image and that the result should return separately, without an awkward success caveat.
Simple requests may be described as taking about a minute. The existing background result and failure paths remain in charge.
Completed experiment cards on Home now give the lead result more space and tuck supporting results into a compact ledger, making history faster to scan without hiding any comparable result.
The homepage now opens with one private health question and evidence-backed answer before showing group challenges, so Murph's one-to-one role is clear first.
Reduced-motion visitors see the complete private exchange immediately, and the group story remains as the second act.
When several shared contact cards use different safe names for the same phone, Murph can now keep up to four explicit alternatives instead of dropping the name entirely.
The alternatives remain unverified display help and never become identity or membership authority.
Eligible new people can start Murph from one iMessage, generated media stays on the private path, and group sharing, sponsorship, recovery, challenge setup, usage, and image requests all have cleaner ways through.
New features
Eligible new people in supported launch markets can text Murph a normal question and receive the answer in the same thread while a full 14-day Pulse trial starts behind the scenes.
The path is limited to direct iMessage and one verified line. Unsupported messages, unsafe admission, or enrollment trouble keep the existing signup-link fallback.
iMessage
In an authenticated group, a participant can ask Murph to share a specific piece of their own private context with that exact room, without setting up a standing group permission.
The request expires within ten minutes, runs through an isolated private read and disclosure review, and grants no future access.
Group chat
An eligible nonmember who likes a group's ordinary join offer now gets one short private opener. Replying returns the phone-bound link for that same group.
Nothing extra is announced in the room, and there is no automated follow-up sequence.
A current participant can make one $5, $10, or $20 contribution, add an optional public alias or note, and turn a larger one-time contribution into a short-lived running bit.
Verified payment adds ordinary cost-weighted group usage credit first. Murph then sends one short original sponsor song in that group; the optional creative moment never controls the credit.
Sponsor this group
Monthly maximumCost-weighted usage credit
Cost-weighted usage credit
Cost-weighted usage credit
Improvements
Generated images and experiment result cards now stay in the member vault and travel through private message attachments instead of public asset links.
Group-avatar updates use a short-lived opaque handoff only for the immediate import. Signed-in experiment participants keep the explicit browser Share or Download flow.
Private attachment
Eligible personal members and Family owners topping up their own seat can ask Murph for the right page, choose $5, $10, or $25, and use one reusable saved card when available.
The amount is never preselected. Card authentication or collection falls back to secure checkout, and only verified payment adds usage.
Settings now combines unused monthly allowance and added or earned usage into one AI usage bar, so a top-up or referral visibly moves the same bar backward.
The recurring reset date stays visible while the internal allowance and credit-source split stays private.
AI usage available
Monthly allowance + added usage
If a recognized member cannot activate Murph in a group because setup, billing, or a trial needs attention, Murph now sends the recovery step privately when a safe direct route exists.
The room never sees account state. Unknown or unsafe cases stay silent or get account-neutral group guidance.
Murph now gives friends the floor when a message is a personal artifact, shared memory, relationship question, or human-to-human beat, even when it ends with a question mark.
A direct ask to Murph still gets a brief answer. If Murph lacks authority to answer a personal fact, it says so plainly instead of guessing or turning the correction into a bit.
Group chat
A loose challenge idea now starts from the group's existing energy and habits instead of a generic intake or an unrequested exercise program.
Concrete games can start immediately, human-owned stakes come first, and photos or cast material stay optional.
Weekly and monthly health outreach now favors verified progress, meaningful context, and useful questions over step-score judgment, tracking guilt, or generic nudges.
Silence is the default when evidence or goal relevance is weak. Zone 2 experiments can also count qualifying walking, cycling, rowing, and elliptical sessions.
If you ask where an image is, Murph now reports whether the original is still generating or queued instead of starting a duplicate.
After delivery, later turns also remember that the prior reply included an image, so Murph does not contradict what arrived.
Ask for a named Murph voice for one voice memo and Murph now resolves that voice correctly, including a save-and-demo in the same turn.
A one-off override does not change your saved voice unless you ask to save it.
iMessage reminders can now land at the overnight time you actually ask for, the iPhone app can recover updated consent without sending you to the web, and long-running group chats keep their maintenance out of the conversation.
Improvements
Ask Murph for an iMessage reminder at 2 AM and it now saves the time you chose instead of warning you away, suggesting a nearby hour, or asking for one more confirmation.
Your own quiet-hour settings and Murph's existing pacing, line-health, consent, and routing safeguards still apply.
Scheduled reminders
If launch consent is missing, the Murph app now presents the current documents and one clear I Consent action, then resumes the setup or sync flow that was waiting.
Home, Settings, legal links, account deletion, and sign-out remain available. Health and automatic meal access pause only when the account has no complete historical consent.
Stay in the Murph app
When a group conversation gets very long, Murph now compacts its working thread during idle time and keeps the synthetic maintenance update out of the room.
The final reply, real progress, errors, and usage accounting still arrive normally. Personal conversations keep their existing in-turn compaction update.
Group Murph can hold onto a room's rhythm, signup and access problems now answer with a way forward, and billing, meal capture, and deletion recover without leaving people stranded.
New features
A group chat can now keep one small, room-local set of recurring bits, preferences, and house rules so Murph does not have to rediscover the room every time.
The current conversation, safety, and explicit room settings always outrank these fallible notes. The group can correct or forget them, and Murph uses a callback only when it naturally fits.
Group chat
After a delivered iMessage invite signup reply, Murph now shares the verified line's native contact card in that direct or group thread so the new member can save it without hunting for it.
The automatic share stays on the line that delivered the reply and remains replay-safe across retries.
Contact card
vCardMurph
Verified line after setup
Improvements
An expired Family invite, an unavailable group connection, a messy Telegram link, or a setup link that was already sent now gets a plain reply instead of looking like Murph ignored the message.
Direct chats can receive the account-specific recovery step. Group replies stay generic so Murph does not disclose anyone's membership or billing state to the room.
When a paused or lapsed member texts Murph from a recognized direct chat, Murph now answers with the existing access or subscription recovery path.
This shared recovery covers billing-access states across eligible iMessage and Telegram direct chats. Suspended accounts keep their existing channel-specific handling.
The group funding page now says how many messages each one-time top-up is expected to add, with the price secondary and the middle option ready to choose.
The counts stay approximate because real model and media costs vary, but every funding surface now derives them from the same estimate.
Keep Murph going
Usage creditCost-weighted credit
Cost-weighted credit
Cost-weighted credit
Paid Pulse and Edge access now derives included monthly AI usage from the recurring plan price instead of a separate fixed allowance.
The included amount is 80% of the catalog price for the member's billing mode and tier. An open period keeps any higher existing amount until renewal, and newly created group chats begin with their own $7.50 included limit.
If you finish the Pulse payment flow and return to Settings in a signed-out browser, Murph now preserves the signed handoff through authentication and resumes it for the right member.
Expired, duplicated, altered, or wrong-member returns stay inert.
An active member who already granted both launch permissions can now turn on automatic meal capture even when a newer legal-document version made the recorded acceptance stale.
No consent or partial consent still fails closed, and every upload continues to recheck current access and the historical grants.
Murph now keeps up to twice as much of a long Grok answer about X and says plainly when the returned result was cut short.
The unverified answer remains isolated from Murph's own instructions, and Murph does not invent missing posts, links, or authors to fill the gap.
If cleanup is interrupted after account deletion begins, Murph now retains the exact remaining work and safely resumes it instead of losing ownership between systems.
Stale billing or access events cannot quietly restore a deleted member while cleanup is still draining.
Before deleting or overwriting something, Murph is now explicitly instructed to verify the exact target and prefer a recoverable path when one exists.
The newsletter, bloodwork, errands, and recovery cards now recompose at 320 to 390 pixels instead of squeezing desktop-sized artifacts into the phone.
Group Murph reads the floor, gets more creative, and knows who is speaking on Telegram. Murph can also ask Grok about X, adapt outdoor reminders to the weather, and close billing and phone-call loops that used to strand people.
New features
Murph can now ask Grok to search X about an account, a topic, or a post you share, then bring Grok's returned answer back into your conversation.
Murph labels the answer as Grok's unverified report and is instructed not to add links or authors beyond what Grok returned. The exact provider cost for the call counts toward your Murph usage with no markup.
A new conversation-only Unhinged dial from 0 to 10 controls how restrained Murph keeps its wording, without changing safety, truth, privacy, consent, tool access, or reminder frequency.
It works in private conversations and group chats with verified members, stays out of onboarding and web Settings, and moves in a small step when you ask for a little more or less. In a group, Murph raises it only when the room's own tone supports it.
When the room earns it, Murph can turn a shared photo into a new chat icon or over-deliver an apology as a song instead of waiting for someone to commission the joke.
The room's own joke is the material. Murph keeps bodies and anyone visibly not in on it out of the bit, changes the icon if someone objects, and stays quiet when the complaint is that Murph spoke at all.
When a linked member writes in a Telegram group, Murph now attributes that turn to the right person and can use their username when no room name exists.
Challenge opt-ins no longer arrive anonymous, so Murph can tell who accepted and keep the group's conversation and shared state attached to the right member.
A reminder that tells you to go outside can now check current conditions and adapt the wording instead of sending you into the rain.
Murph uses only a city or region you already shared, or offers once to save one. Declining does not block the reminder, and a failed weather read never prevents it from going out.
Improvements
Murph now distinguishes a live volley from catch-up, a joke from a plausible emergency, and a human-to-human beat from an opening that actually wants Murph.
It may wait a few seconds while people are mid-burst, gives human punchlines more stage, and asks one short question when the health context is genuinely unclear. Timing never overrides the room's silence, reaction, or not-for-you rules.
When a call Murph started finishes, Murph now sends the meaningful result back through the chat where it can reach you, in its own voice.
The result is encrypted at rest, Murph treats the call transcript as untrusted, and a replay cannot send the same outcome twice.
If your subscription lapses, Murph answers on your home line and points you to Subscription controls instead of dropping the text or sending you to a blocked signup wall.
Resuming a paused subscription now carries the saved card onto the invoice, avoids an update the provider rejects while paused, and offers the billing page when the provider is still finishing.
After sign-in, accepting Murph's terms and health-data notices now takes one affirmative action instead of a row of checkboxes, with an equally clear Decline path.
The prompt states that Murph does not sell your health data or train general-purpose AI models on Murph-managed health data. After repeated save failures, it offers a support route instead of trapping you in retries.
When a group runs out of included usage, Murph now explains the pause in its own voice and, when funding is available, gives the whole room a clear way to bring it back.
On iPhone, Add Murph to Contacts now hands off from an in-app browser to Safari instead of letting the embedded window swallow the contact-card download.
The short-lived handoff stays bound to your account and chosen avatar. Setup completes only after Safari actually opens; if the launch is blocked, the picker stays open with a retry.
A leftover folder in experiment storage can no longer disable every managed automation or archive active experiment check-ins. Unrelated reminders keep running while Murph retries a temporary experiment scan failure.
Murph can now rename the chat it is already in even when the group has not started a challenge or created a shared workspace. Updating Murph's own label follows when possible.
Murph now receives a safe description of why voice generation failed, while the maximum script length and generation timeout share one limit so overlong memos do not fail by design.
Provider status and request identifiers reach secret-safe logs, giving Murph and support enough information to distinguish a timeout from a quota or request error.
If a database timeout proves that no query or transaction started, Murph's web service now retries it briefly instead of failing the request.
A failure that may have reached the database is never replayed, so the retry cannot duplicate a completed action.
The iMessage roster now separates people who completed Murph setup from people who never activated, without treating current plan status as identity. A paused or lapsed member is no longer mistaken for a stranger.
If a push-based device source still looks connected but stops delivering data, Murph now records the silence as a source-stalled signal instead of treating every empty sync as a normal empty day.
The check is observation only: it does not disconnect the source, block ingestion, or launch a recovery on its own.
The hero now tells one story from the start: friends enter the chat, the challenge begins, and Murph carries it through standings, a roast voice memo, and the Sunday recap.
The separate solo-demo act and its decorative controls are gone, and the security promises now read in one clean column.