Hi friend,

There’s lots to cover and I know you’re busy.

So let’s get into it.

⏺️ Blossom Health raised $20M to deploy AI copilots in psychiatry practices across the US. The funding targets reducing wait times and administrative burden in psychiatric care through AI-powered clinical decision support. (link)

⏺️ Amsterdam-based OpenUp raised €20M to expand its employee mental health platform across Europe. (link)

⏺️ JAAQ raised a $17M Series A. The London-based company has a huge library of clinical content that its clients (including employers, insurers, and health systems) can license for use in their own products. The funds will be used primarily to support US expansion.

⏺️ Gilgamesh Pharma raised a $60M Series A to develop next-generation neuropsychiatric therapies. (link)

⏺️ Cerebral acquired ADHD app Inflow. The deal pairs Inflow's D2C ADHD business with Cerebral's telepsychiatry, medication management and teletherapy services (thanks to their Resilience Lab acquisition). Deal terms were not disclosed. (link)

⏺️ Otsuka Pharmaceutical is acquiring Transcend Therapeutics for $1.225B. Transcend is a clinical-stage neuropsychiatric company with a pipeline focused on rapid-acting treatments, particularly for PTSD. (link)

⏺️ Vitality acquired Ramp Health. The deal will to combine Vitality’s AI-powered health behaviour change platform with Ramp's onsite clinical and coaching services. (link)

⏺️ Sword Health announced the “Chapter II” of their mental health product, Mind. Sword are betting heavily on chat as the interface for care, with Mind now a three way chat between a client, their doctoral level psychologist and Phoenix, their AI Care Specialist. Sword also announced that they have built their own foundational model for mental health, which powers Phoenix. It looks like Sword are trying to build the killer-app for the employer wellbeing market, positioning Mind as a full EAP, with a single interface, that addresses all aspects of an employee’s wellbeing. They also just announced that the business surpassed $400M in ARR, adding over $100M in ARR in the last two months, which is absolutely wild! (link)

⏺️ Slingshot launched Ash Align, a two-person AI conversation tool. The feature lets two people each speak privately with Ash before then being brought into a shared conversation. It will be interesting to see if this gets consumer traction. (link)

⏺️ Mantra Health launched Beacon, an early alert and analytics product. Beacon will allow higher ed institutions to better identify risk in their student groups and make more data informed decisions in how they manage student mental health. (link)

⏺️ The University of Cambridge launched the Mental Health Catalyst, an initiative to accelerate innovation in depression treatment. Led by the Institute for Manufacturing and IfM Engage and backed by the Wellcome Trust, the programme will convene researchers, clinicians, funders, and people with lived experience to develop a Mission Map for coordinating research and investment. (link)

⏺️ A California jury found Meta and YouTube liable for designing their platforms to be addictive, awarding $6M in damages to a plaintiff who said social media worsened her depression and anxiety as a child. In the same week, a separate New Mexico jury ordered Meta to pay $375M for concealing child safety risks. (link)

⏺️ Apple's App Store introduced a requirement for health apps to disclose their medical device regulatory status in the US, UK, and Europe, with compliance required by early 2027. (link)

⏺️ US mental health and substance use disorder spending grew from $40.9B in 2000 to $139.6B in 2021. Mental health and SUD spending grew from 4.5% to 5.5% of all medical services spending over that period, at an average annual real per capita growth rate of 3.27% - faster than overall medical spending at 2.21%. 87.3% of the growth was driven by more people receiving treatment rather than rising costs per case. (link)

If you find these updates valuable and would like even deeper insights on mental health innovation, consider becoming a Hemingway Pro member.

As a member, you’ll join a community over 400 mental health leaders and get access to exclusive content, events and resources. We have in-person events in SF, Chicago Sydney with more coming to London, Chicago and DC soon. We are also hosting online, expert led events with leaders like Headspace CCO Jenna Glover, Dana Udall PhD and more.

Research News:

⏺️ Agentic AI system approaches diagnostic accuracy of board-certified psychiatrists. Researchers published WiseMind, a multi-agent AI framework for psychiatric diagnosis, in npj Digital Medicine. The system works by running two AI agents in parallel - one handling clinical reasoning guided by a DSM-5 knowledge graph, the other handling empathetic communication. These agents conducted diagnostic interviews across 1,206 simulated and 180 real user sessions and achieved 85.6% top-1 diagnostic accuracy. The authors report this approaches the performance range of board-certified psychiatrists and outperforms other LLM-based approaches by 15–54 percentage points. (link)

⏺️ Spring Health shared results of their AI product on their own VERA-MH benchmark. It’s pretty cool to see companies publicly sharing results like this on an open source benchmark that anyone can use to test their own products. (link)

⏺️ John Torous published a post discussing how some of the evaluation methods for AI are insufficient and may be misleading. It echoes some points we discussed in our recent deep dive on AI evals. (link)

⏺️ A machine learning model combined 103 biomarkers with demographic and clinical data to predict suicide risk among psychiatric inpatients. The study analysed 2,785 first-admission patients across three institutions, with suicide risk assessed one week after admission. The model achieved an AUC of 0.808 in predicting suicide risk - which is strong. Nine biomarkers were associated with elevated risk and six with potentially protective effects. The study is observational and the model has not been tested in clinical deployment. (link)

⏺️ TabR, a retrieval-augmented neural architecture, predicted suicide attempts in Korean adolescents with 97% accuracy. (link)

⏺️ According to a new study of 4,578 participants, apathy, depression, and anhedonia are distinct conditions that can be reliably told apart using just 10 symptoms. Factor analysis identified five separate dimensions - including three subtypes of apathy - with machine learning able to distinguish between them with greater than 90% accuracy. (link)

⏺️ “Internet-based Problem Management Plus” (IPM+) produced significant improvements in depression, anxiety, and stress among pregnant women in a randomised trial. The five-week digital intervention achieved moderate-to-large effect sizes. (link)

⏺️ Array Behavioral Care published outcomes data from 5,805 patients. Array achieved 27% remission and 60% response, compared to benchmarks of 16–22% remission and 46–47% response from large integrated health systems. The study did not have peer review and was published by Array. (link)

⏺️ Human-AI interaction research lacks the psychological theory needed to produce durable findings, according to a Nature Reviews Psychology paper. Without theoretical frameworks that outlast any particular model release, the authors argue that findings in the field risk becoming quickly obsolete, and that existing social cognition theory offers a more durable foundation. (link)

⏺️ UNICEF Innocenti published a report authored by 30 adolescent co-researchers from seven countries on what children want from digital governance. Their core asks: mandatory AI content labelling, stronger privacy protections, and inclusion of children as co-designers of digital policy, not just as subjects of it. (link)

That’s all for this week.

Keep fighting the good fight!

Steve Duke

Founder of Hemingway

P.S. Find this valuable and want to support our work? Consider becoming a Hemingway Pro member.