What this is about:
Meta’s new Muse Spark model is being pitched as a more capable, multimodal assistant — one that can reason about questions in “science, math, and health,” analyse images and charts, and eventually sit across Meta’s full ecosystem, including its AI glasses.

Why it matters:
Health is not a side quest. Meta explicitly says that health is one of the top reasons people turn to AI, and showcases use cases such as estimating calories from food images and answering health questions using photos and charts. Thus a shift, from chatbot as convenience tool to chatbot as quasi-health interface.

What’s happening:
Meta says it worked with physicians to make Muse Spark more helpful on common health questions. What it has not published, at least in the launch materials, is the kind of detail clinicians would actually want:

In other words, this is a product launch wearing a health halo.

The health problem:
The combination of multimodal input plus consumer trust is where this gets interesting. If users start uploading lab results, rashes, meal photos, sleep data or symptom screenshots, the tool starts to resemble a health triage surface, whether Meta wants that label or not.

Wired’s early test was not encouraging: Muse Spark reportedly asked for raw health data and then produced poor advice.

Reality check:
There is a real science base behind digital phenotyping and wearable-derived health signals. The field is not fake. Research in npj Digital Medicine and npj Mental Health Research shows that densely sampled data from phones and wearables can generate clinically useful signals, including around mood, sleep and depressive episodes.

But those same papers make the harder point too: the challenge is not collecting more data, it is validating which signals actually map onto mental states and outcomes before deploying them at scale.

The critique:
Meta is asking the public to trust a health-facing AI product before showing the kind of evidence that would normally earn that trust.

That is a particularly uneasy fit in a field where safety and robustness remain unresolved. A 2025 JAMA Network Open study found commercial LLMs highly vulnerable to prompt-injection attacks in medical contexts, with unsafe recommendations generated in 94.4% of simulated trials. A separate JAMA Network Open paper found that patients place more trust in medical AI when clinician oversight, representative data, and visible governance are built in — precisely the conditions least apparent in a consumer AI launch.

The interesting bit:
The deeper issue is not whether Muse Spark can answer some health questions reasonably well. It probably can. The issue is whether consumer AI companies are normalising a new layer of health inference before the governance, validation and transparency are mature enough to support it. In mental health especially, that should ring alarm bells.

The line:
Meta has not launched a doctor. It has launched a health-shaped reason to hand over more intimate data.

Sources: Meta launch post / product page | WIRED | JAMA Network Open: prompt-injection in medical LLMs | JAMA Network Open: patient trust in medical AI | JAMA Network Open: CHART reporting guideline | npj Digital Medicine: digital phenotyping | npj Mental Health Research: digital biomarkers in bipolar depression | The Lancet Digital Health / series paper

I’m watching CNN because, frankly, British news bores me sh*tless, and the headline pairing of “the iconic American Girl dolls are back” with “strait showdown”, I nearly choked on my Babybel. War scrolling beneath first-world nostalgia. The Pope gets targeted, and my mum rings in a state of hysteria to announce that “Trump has turned himself into Jesus.” You literally couldn’t write this sh*t. Meanwhile petrol prices climb, the cost of living keeps squeezing people, the drum of unsafe uncertainty beats on.

My patients are frightened, dysregulated and overexposed. I’m not asking them to ignore what’s happening, but I am asking them to stop bathing in it. We could all do with less cortisol. There is a difference between being informed and marinating in catastrophe, and right now far too many people are confusing the two. The nervous system does not care whether the threat is on your street or on your screen. It just knows it is bloody overwhelmed.

I spent the weekend on the doors as usual, immersed in residents’ casework and the texture of actual life. That’s all I have got really, when the world goes grandiose and grotesque, come back to what is in front of you. I am not here to berate anyone for their political views. I have my biases and I do not hide them, but that does not mean I shun or shy away from people whose views differ from mine. I can’t, can I? If you want to work with people, truly work with them, you cannot start by deciding that everyone who thinks differently from you is morally dead to you.

What I do know is that I despise the amount of time I am having to spend on social media in the run-up to this election. It is affecting my mental health. It is a mood hoover. Yesterday I felt completely numb, so I put my phone into grayscale mode to try to step off the addiction pathway. Grim, but effective. If a few months of this, and it not even being my actual job, feels this mentally damaging, I genuinely wonder what state influencer mental health is. Overwhelmed and overstimulated. No amount of money in the world would persuade me to live at that pitch permanently.

I did manage some escape: to Oxford for a candlelit Vivaldi Four Seasons at the Holywell Music Room: the oldest purpose-built music room in Europe. Which sounds insufferable until you’re actually in it and realise the history does half the work for you. Low light, old wood, strings, no phones, no pundits. My brain emptied on contact. Then we went to our friend’s Art gallery, Oink, where Donk was in residence (lovely chap), and Max bought me another piece for my office: come here and don’t think about it. #Accurate.

Over the weekend, I spent time at the ACA Conference in Columbus, Ohio, speaking with hundreds of counselors at the Blueprint.ai booth and handing out therapist care packages.

What stayed with me: Therapists rarely stop to acknowledge the positive impact of their work.

There is so little time to pause, and so many other things competing for attention, even though this is something we ask of our clients all the

This week’s tool: Take a moment to recognize the good you do.

Maybe that means noticing:

Something to think about: Being surrounded by thousands of counselors made the ripple effects of mental health professionals feel visible.

If each therapist in that space has helped, and will continue to help, countless people, the collective impact is hard to fully grasp.

And all of that happens while managing paperwork, emotional weight, and the many demands that exist outside the therapy hour.

I left feeling proud, humbled, and genuinely grateful to be part of this profession. It also felt meaningful to have a therapist care package to offer in return.

If you’d like to send a free care package to yourself or to a therapist you know, here’s a link to the Therapists Deserve Better gifting campaign.

What happened: A preregistered longitudinal study in npj Mental Health Research examined whether people with a greater general tendency to forgive others reported better well-being about a year later. The sample was very large: 207,919 participants across 23 countries in the Global Flourishing Study.

A notable stat: 75% of participants said they “often” or “always” forgave people who hurt them. But that varied widely by country, from 41% in Turkey to 92% in Nigeria.

What they found: Higher forgivingness was associated with better later well-being across multiple domains, though the effects were mostly small. The strongest links showed up in psychological well-being, especially outcomes like optimism and meaningful activities.

Cultural context matters: The cross-country variation suggests forgiveness is not just an individual trait. As the study’s lead author, Richard Cowden, wrote in The Conversation, rates of forgiveness appear to be "shaped by cultural and contextual influences, including social norms around harmony and religious teachings about wrongdoing.”

Why it matters: This suggests forgivingness may play a role in flourishing, but not in a sweeping or universal way. A lot depends on the cultural context in which the person lives.

Therapist takeaways:

The show is stressful, yes, but importantly, it’s also very real. Season 2 has been excellent - same intensity, different day (If you’re not aware, each season takes place over the course of one day).

There are characters I love (Dr. Mel King, Dr. Robby, and nurse Dana) and at least one I would never want as a co-worker - looking at you, Dr. Trinity Santos. If you’re not watching The Pitt yet, consider this your not-so-gentle nudge to jump on board.

I have been listening to The Four Seasons on repeat since the concert, and the science fully supports why it feels like such an effective head-emptier: multiple brain systems engaged in emotion, memory, attention and prediction, which goes some way to explaining how it can steady the mind and shift one's internal weather so quickly.

There is even research on the so-called “Vivaldi effect”.

So yes, I am prescribing it accordingly alongside noise-canceling headphones and, ideally, a purring cat on your lap. Not a cure for anything, but very decent nervous system admin.

Stories from the community

The clear winner is authenticity and intentionality, even if a blank wall is the only usable space in our home.

Notes from the Community: