Hi friend,

I’m back in London after a hectic but wonderful week in Nashville. To those I saw at BHT, thanks for making it so fun. For those I didn’t see, we need to do an IRL catchup soon - it’s just the best!

Our round-up is a little late this week… I decided it would be a good idea to move house during the busiest week of my year… so I hope you’ll excuse me. Better late than never, as they say!

The good news is that it was a super interesting week in mental health. We had two large, interesting fundraising announcements with Kasvu Therapeutics and ReEmerge. Pelago and Nosis had big product announcements and Kooth released their H1 earnings.

Did you think we had enough mental health AI benchmarks? Well, think again, my friend! OpenAI released MentalHealthBench, and Spring added a new framework to VERA-MH to evaluate how AI models perform when a user may be at risk of physical or sexual harm from another person — I actually found this one novel and interesting.

On the research front, an interesting study found that a machine learning algorithm was unable to predict who would benefit from a digital, single-session intervention for depression. An early trial found Alkermes’ experimental ADHD drug shows promise, and a study from Boehringer Ingelheim and Click Therapeutics showed that CT155, their digital therapeutic, modestly reduced motivational symptoms of schizophrenia

Let’s get into it!

Want to go to Inside the Lab by APA Labs?

Next month, APA Labs is hosting the Inside the Lab event in Austin. It's a hands-on, collaborative event (think more workshops than speeches) where a variety of stakeholders come together to address the tensions and challenges of innovating in mental health. A lot of Hemingway community folks went last year, and I heard really great feedback — many are returning this year.

We're partnering with APA Labs again this year to offer a 20% discount to Hemingway Readers who would like to go. Just use the code ITL2026WFOOY when signing up.

What happened in mental health last week?

Industry News:

⏺️ Kasvu Therapeutics raised €30M to develop a non-hallucinogenic treatment for depression. The Finnish biotech is developing KTX-0141, a drug designed to enhance TrkB signalling and neuroplasticity without activating the serotonin receptor responsible for psychedelic effects. The Series A will take the drug through studies needed to enter human trials and into a planned Phase I/1b trial in major depressive disorder beginning in the second half of 2027. (link)

⏺️ ReEmerge raised $37M to test deep-brain stimulation for lasting problems after traumatic brain injury. The company is developing an implanted brain-stimulation treatment for people who continue to experience problems with attention, memory and other cognitive functions long after a moderate-to-severe brain injury. The company emerged from stealth with the Series A co-led by U.S. Venture Partners and Santé. (link)

⏺️ Pelago launched its behavioural health platform. The company, which previously focused primarily on substance-use disorders, is expanding its services to include treatment for conditions such as anxiety and depression, as well as behavioural addictions including gambling, gaming and binge eating. People entering the service will first interact with Sona, Pelago’s voice-based AI, which gathers information about their needs before they are directed to appropriate care, including clinicians where needed. The platform is positioned as “acuity-matched care”, an attempt to address payers’ concerns over rising behavioural health costs. The platform will become available to participating employers and health plans in November. (link)

⏺️ SUD Platform Nosis came out of stealth and launched across all 50 US States. The startup offers a whole-body health model for substance use recovery. This includes an AI-supported assessment that examines more than 85 biomarkers alongside health history and symptoms, as well as a 12-week programme focused on nutrition, exercise, sleep, stress and metabolic health. Members work with health coaches who are themselves in recovery, as well as registered dietitians. Nosis was part of the 2025 One Mind Accelerator cohort and will be an interesting company to watch in 2026 and beyond. (link)

⏺️ Kooth announced their H1 results, doubling adjusted EBITDA despite a 3.9% fall in revenue. The digital youth mental-health company reported £30.8M in revenue for the first half of 2026, down 3.9% year on year, while adjusted EBITDA rose 102% to £5.3M and the company moved from a £1.3M loss after tax to a £2.0M profit. Kooth said its services now cover 20.2 million people across California, Michigan, New Jersey and the UK. Attention remains heavily focused on the performance of their Californian contract and the potential for its continuation. Kooth’s share price is up 63% since the start of the year. (link)

⏺️ OpenAI released MentalHealthBench to test how AI responds to mental-health conversations. The benchmark gives AI models synthetic conversations modelled on real-world interactions, ranging from relationship problems and everyday distress to serious mental-health concerns and emergencies. It then asks the models to respond to the final message. More than 80 psychologists and psychiatrists from 22 countries reviewed the scenarios and created specific criteria for what the AI should and should not do in each one — for example, asking for important missing information or supporting a user’s own decision-making, while losing points for potentially harmful responses. An automated grader then scores the AI response against these expert-written criteria, allowing researchers to compare models and identify where their responses fall short. It is the latest in a proliferation of mental health AI benchmarks and is publicly available for other researchers and developers to use. The release from OpenAI also includes results from recent frontier models tested against the benchmark. (link)

⏺️ Spring Health released a framework for testing how mental-health AI responds when someone may be experiencing violence or abuse. The new rubric extends VERA-MH, Spring Health’s framework for evaluating the safety of AI used in mental-health settings, to situations where an adult may be at risk of physical or sexual harm from another person. Like all benchmarks, testers can present an AI system with these scenarios and assess its response against specific criteria. These criteria include whether the model recognises signs of danger, asks enough questions to understand the immediate risk, responds supportively without blaming the user, and directs them towards appropriate human or emergency support when needed. Spring Health has released the draft rubric for public comment. (link)

⏺️ Rula announced an external council to review its use of AI in behavioural health. Rula “CARES” brings together clinical, academic and AI experts to provide outside review as the company develops AI-enabled products. Rula says the council will examine issues including clinical safety and ethical implementation. (link)

⏺️ Georgetown’s Thrive Center selected seven organisations for its mental-health innovation fellowship. The three-month programme will support organisations working on child and family wellbeing, including AI mental-health platform Dr. Katz, FatherLink AI, paediatric therapy company Ladder Health and Understood. (link)

⏺️ Cambridge’s Mental Health Catalyst announced partners for their programme to identify, develop and encourage the uptake for promising interventions for depressed mood. The Mental Health Catalyst, a major programme led by the University of Cambridge’s Institute for Manufacturing and supported by Wellcome, will bring together the William Templeton Foundation for Young People’s Mental Health, Cochrane’s Evidence Synthesis Unit, Generation Mental Health Association and the Mental Health Innovation Network. The programme aims to identify promising interventions for clinically significant depressed mood and tackle the barriers preventing them from being developed and adopted. (link)

⏺️ UCL and the Global Brain Economy Initiative formed a brain-capital research partnership. The organisations will create a working group linking research on brain health and skills with economic policy, and plan to convene governments, researchers and other organisations through UCL’s New Alignment Hub. The partnership is intended to translate the emerging concept of “brain capital” into research, measurement and policy rather than test a specific health intervention. The Brain Economy movement continues to gain traction and is something I am watching closely. (link)

⏺️ The FDA held a public hearing on the future use of psychedelics in healthcare. Reporting from Psychedelic Alpha noted that roughly 80 speakers gave the FDA and other federal agencies their views on how psychedelic treatments should be regulated. Discussion focused on practical questions that could shape future approvals, including whether psychotherapy should accompany psychedelic drugs, how providers should be trained, what safety restrictions should apply and how treatment outcomes should be monitored. Speakers differed on several of these issues, although most were broadly supportive of therapeutic psychedelic use. The hearing was held to gather public input rather than make regulatory decisions, and federal officials did not question speakers or announce any changes in policy. (link)

⏺️ Some NHS bodies introduced minimum waits of up to two years for ADHD and autism assessments. Four integrated care boards in England have introduced two-year minimum waits in parts of their assessment pathways, while 15 ICBs covering 19 million people are using some form of restriction as demand and spending rise. The wait times were introduced as a way to “ration” assessments for ADHD and autism as services strugfle to meet demand. England currently has about 960,000 people waiting for an ADHD assessment and 295,000 waiting for an autism assessment, according to the Guardian’s reporting. (link)

Join more than 500 other mental health leaders who are Hemingway Pro members. As a Pro Member, you get access to deep industry insights, access to our private Slack group and invites to our member-only events.

Research News:

⏺️ A study found demographic information changed how AI diagnosed otherwise identical eating-disorder cases. Researchers gave ten configurations from five large-language-model families 1,152 synthetic clinical cases in which symptoms were held constant while characteristics such as race, gender and age were changed. In deliberately ambiguous cases, Black patients were six times as likely and Latine patients more than nine times as likely as otherwise identical White patients to receive a depression diagnosis, while Asian patients were nearly three times as likely to receive an anorexia diagnosis. The experiment used synthetic cases rather than real patients, but shows that demographic information alone can alter diagnostic outputs. (link)

⏺️ A study found a machine learning algorithm could not reliably identify who would benefit most from a single-session digital depression intervention. Researchers revisited a trial of 828 adults testing COMET, a 25–40 minute online intervention teaching skills drawn from cognitive behavioural therapy and positive psychology. COMET had not reduced depression more than a waitlist overall, so the researchers tested whether machine learning could instead identify particular people who would benefit, using information including their symptoms, sleep, social support, childhood experiences and personality. The model could not reliably predict who would fare better with COMET than with no intervention. The findings suggest that the patient characteristics measured in the study were not enough to determine who should receive this particular intervention. While personalised treatment remains a major goal of mental health care, it continues to prove elusive. (link)

⏺️ A randomised trial found a digital therapeutic modestly reduced motivational symptoms of schizophrenia. Researchers assigned 457 adults with schizophrenia to CT-155, a smartphone-based therapeutic used alongside antipsychotic medication, or a digital control for 16 weeks. Symptoms involving motivation and pleasure — such as difficulty pursuing goals, socialising or enjoying activities — improved by 6.8 points with CT-155 compared with 4.2 points in the control group, a 2.6-point difference on the CAINS-MAP scale. The treatment did not significantly improve positive symptoms such as hallucinations or delusions. The study was funded by CT-155’s developers, Boehringer Ingelheim and Click Therapeutics. (link)

⏺️ A small trial found psilocybin therapy was generally well tolerated and showed preliminary benefits for bipolar II depression. Researchers gave 14 people experiencing bipolar II depression a 10mg dose of psilocybin alongside psychotherapy, with a 25mg dose offered to those who remained depressed. Depression scores fell after treatment, with four participants entering remission after the initial 10mg dose and further improvements observed among those who received 25mg. Most adverse effects were mild or moderate, although three participants experienced psychiatric events involving suicidal thoughts or hypomania that resolved with support. The study had no placebo group and only 14 participants. The authors concluded that psilocybin may be a safe, tolerable and potentially effective treatment for treating depression in BD-II, but randomised controlled trials are needed to confirm efficacy. (link)

⏺️ An early trial found Alkermes’ experimental ADHD drug was well tolerated and showed preliminary improvements in symptoms. In the Phase 1b study, 50 adults with ADHD stopped their existing medication and were randomly assigned to receive one of two doses of ALKS 7290 or placebo for 14 days. ALKS 7290, which targets the orexin system involved in wakefulness and attention, was primarily being tested for safety; exploratory measures showed ADHD symptom scores fell by a median 14 points with the 20mg dose and 19 points with the 50mg dose from baseline. However, the study was small and was not designed to determine whether these improvements were significantly greater than placebo, so efficacy remains uncertain. No serious treatment-related adverse events were reported, and a 312-person Phase II trial is now underway to test efficacy against placebo. (link)

⏺️ A study found people with schizophrenia described difficulties that extend beyond standard symptom categories. Researchers synthesised qualitative studies of people describing their own experiences of schizophrenia-spectrum disorders and organised these accounts into a framework of recurring experiences. Alongside familiar symptoms such as hallucinations and reduced motivation, people described more fundamental changes in their sense of self, connection with other people and the world around them, as well as difficulties making sense of their experiences. The authors argue that these first-person experiences are not fully captured by conventional clinical measures focused on positive, negative and cognitive symptoms, and could provide a broader picture of how schizophrenia affects people’s lives. (link)

⏺️ An international group argued that loneliness among young people requires a broader public-health response. Writing in World Psychiatry, the authors review evidence linking loneliness in children and adolescents with depression, anxiety, suicidal thoughts and poorer physical health, and argue that it has received less attention than loneliness in older adults. The burden is particularly high in some low- and middle-income countries: data cited in the paper suggest around 24% of adolescents in LMICs experience loneliness, compared with about 11% in high-income countries, while almost 90% of the world’s adolescents live in LMICs. The authors call for better measurement of loneliness across cultures and more research into prevention and interventions, particularly approaches that can be delivered through schools and communities rather than relying solely on mental-health services. (link)

That’s all for this week.

Keep fighting the good fight!

Steve Duke

Founder of Hemingway

P.S. If you find these updates valuable, consider becoming a Hemingway Pro Member. As a Pro Member, you will get access to even deeper industry insights and will be invited to join our community of over 500 mental health leaders.