One of Rhode Island’s most-wanted fugitives — a former doctor convicted of sexual assault but on the run for 20 years — appears to have been living a secret life as a biotech executive.
Ronald Fischer, 70, was arrested last week by federal and Rhode Island authorities after they tracked and boarded a 56-foot sailboat cruising off the coast of New Jersey. The former anesthesiologist disappeared in 2005 while on trial for first-degree sexual assault.
The boat was registered under the name Richard Graydon, an alias used by Fischer, U.S. Marshals said. That is the same name as a doctor and seasoned drug development executive hired last March by Immix Biopharma, a Los Angeles-based biotech company, as its new chief medical officer.
Good morning, everyone, and welcome to another working week. We hope the weekend respite was relaxing and invigorating because that oh-too-familiar routine of meetings, deadlines, and the like has returned with a vengeance. You knew this would happen, yes? To cope, we are relying, as always, on cups of stimulation. Our choice today is laced with traces of cocoa. Feel free to join us. Remember, no prescription is required. Meanwhile, here are some tidbits to help you along. Best of luck accomplishing your goals today and, of course, do keep in touch. …
Novo Nordisk, maker of Ozempic and Wegovy, lists the telehealth company LifeMD on its website as a provider that offers “legitimate medicine sourcing and patient support” for people seeking GLP-1 drugs. But some former employees describe LifeMD differently, as a company that has sought to maximize the volume of prescriptions it doles out at the expense of patient safety, STAT reports. Former workers told STAT that providers were pressed to expedite their work to a pace that was not clinically responsible, with two of them saying providers at times were expected to review the cases of 25 people per hour based only on electronic forms the patients filled out — the equivalent of spending about two minutes on each case.
A recent audit of Iowa Medicaid records revealed how pharmacy benefit managers are using complicated and sophisticated approaches for handling prescription drug claims that ultimately overcharge taxpayers, a finding that underscores controversy surrounding these crucial middlemen in the pharmaceutical supply chain, STAT explains. The audit found that one large pharmacy benefit manager appeared to have made more than $100 million by adjusting the amount of money paid to pharmacies without passing some of the savings back to managed care plans working on behalf of the state, according to Iowa officials. The review scrutinized records from 2019 through 2021.
Get your daily dose of health and medicine every weekday with STAT’s free newsletter Morning Rounds. Sign up here.
Good morning. Yesterday’s game wasn’t the most exciting, but luckily the third place game on Saturday had all the drama I’d hoped for. Before replaying the highlights, catch up on a bunch of news below.
This is today’s edition of The Download, our weekday newsletter that provides a daily dose of what’s going on in the world of technology.
AI is more likely than humans to form biases when hiring
The next time you apply for a job, AI may screen your résumé before any human sees it. But there’s good reason to question whether AI will judge you fairly.
We already know that LLMs pick up human biases from their training data. New research suggests they can also develop their own biases from experience—and stereotype job applicants more than humans do.
As AI companies race to build agentic models that remember the tiniest details about users, they may be handing them ammunition for forming those biases.
Every morning, airline dispatchers, grid operators, and farmers around the world make decisions based on weather forecasts. More recently, the forecasts have become relevant for another industry: prediction markets, where people bet money on all kinds of real-world events, including the weather.
The temptation to manipulate weather data to get an edge in these markets, combined with a collective move toward data-driven AI weather forecasting, is starting to put the accuracy of weather predictions at risk.
As experts in the field, we can foresee scenarios where the risks snowball into far bigger, more systemic problems.
—Monique Kuglitsch, Jesper Dramsch, Franz G. Kuglitsch, & Andrea Toreti
The must-reads
I’ve combed the internet to find you today’s most fun/important/scary/fascinating stories about technology.
1 SpaceX is negotiating to sell the Pentagon AI compute It would provide data center capacity worth billions of dollars. (WSJ $) + Deepening ties between Elon Musk’s company and the DoD. (Reuters $) + Meanwhile, Anthropic is in talks with Meta to acquire compute. (CNBC) + The compute explosion is only just beginning. (MIT Technology Review)
2 Trump Media wants $100,000 a month for early access to Trump’s posts The premium feed is being pitched to trading firms and banks. (FT $) + It aims to monetize Trump’s market-moving social media posts. (Reuters $) + Critics described the plan as “brazen corruption.” (Guardian)
3 ICE shared Medicaid data it wasn’t supposed to have with Palantir Court filings show the data reached the contractor before being deleted. (NPR) + ICE is using data broker tools to identify “unaccompanied minors.” (Wired $)
4 Apple briefly overtook Nvidia as the world’s most valuable company The iPhone maker’s earnings durability has impressed investors. (Reuters $) + While Nvidia’s rise has stalled amid shifting AI bets. (CNBC)
5. Politicians are trying to change what chatbots say about them A new industry has sprung up to help them edit AI outputs. (NYT $) + Chatbots can sway voters better than political ads. (MIT Technology Review)
6 Washington is opening the door to armed robots The Pentagon is accelerating AI weapons development. (WP $) + “Humans in the loop” in war is an illusion. (MIT Technology Review)
7 China’s Moonshot has paused new subscriptions amid surging uptake Demand for the headline-grabbing Kimi K3 has strained capacity. (SCMP) + China’s open-source AI is challenging US models. (MIT Technology Review)
8 Lab-grown teeth could soon replace fillings and implants Scientists believe regenerative medicine could transform dentistry. (BBC) + Humanlike “teeth” have been grown in mini pigs. (MIT Technology Review)
9 AI slop on birdwatching forums is putting research at risk It could contaminate records of species. (Guardian)
10 Heart experts have good news for your coffee habit Roughly five cups per day is fine—and may even be beneficial. (Gizmodo)
Quote of the day
“The most authoritarian government is producing the most egalitarian models, and what should be the most democratic government is breeding companies that are the most authoritarian.”
—Rayan Krishnan, CEO of Vals AI, a company that evaluates AI performance, gives the New York Times his take on the competition between Chinese and American models.
One More Thing
RICHARD CHANCE
The curious case of the disappearing Lamborghinis
A new wave of theft is rocking the luxury car industry—mixing high tech with old-school chop-shop techniques to snag vehicles while they’re in transport.
It’s remained under the radar, even as it’s rocked the industry over the past two years. MIT Technology Review identified more than a dozen cases involving high-end vehicles, obtained court records, and spoke to law enforcement, brokers, drivers, and victims in multiple states to reveal how transport fraud is wreaking havoc across the country.
Modern deep learning has broadened the tools available for non-invasive neural decoding, but its advantage over well-engineered classical pipelines remains unclear at clinical neural-engineering sample sizes. We compared four classical decoders, three multilayer perceptron (MLP) variants, and three deep-learning architectures (an EEGNet-style compact convolutional network, a four-layer Transformer encoder trained from scratch, and a graph attention network) on the public WAY-EEG-GAL grasp-and-lift dataset (12 participants, 3,528 trials). Models were evaluated using leave-one-subject-out (LOSO) cross-validation to decode object weight (165, 330, and 660 g) and grasp-surface friction (sandpaper, suede, and silk). After Benjamini–Hochberg false discovery rate (BH-FDR) correction within the primary/robustness family, none of the deep-versus-best-classical comparisons reached significance. The graph attention network led nominally on weight (0.643), and the compact convolutional network (CNN) led nominally on surface (0.565), but neither exceeded the best classical baselines (HGBM = 0.617 for weight; logistic regression = 0.562 for surface). Two-direction bandwidth controls showed that the nominal graph neural network (GNN) advantage on weight reflected access to higher-frequency electromyography (EMG) content rather than a robust architectural gain. The GNN weight signal was concentrated in the first 500 ms of sustained hold (0.639 early vs. 0.540 late, q = 0.007), whereas the compact CNN was comparatively robust to bandwidth and phase. A four-layer Transformer trained from scratch underperformed (q = 0.003 for both tasks), consistent with a parameter–data mismatch at n = 12. Modality ablation showed that both tasks were dominated by peripheral EMG features: electroencephalography (EEG)-only decoding was near chance for weight (0.34–0.37) and surface (0.36–0.38; all q < 0.0015 vs. EMG-only). EMG-only HGBM achieved the highest accuracy in the study (0.712 for weight and 0.584 for surface), and adding EEG channels reduced HGBM weight decoding (q = 0.007) and CNN surface decoding (q = 0.012). In contrast, the graph attention network was robust to EEG-induced fusion dilution (q > 0.5), consistent with attention-based down-weighting of low-information channels. Together, bandwidth, modality, phase, and conditioning controls indicate that, at clinical sample sizes, the critical design choice is not architectural complexity but modality-relevant channel selection and the ability to suppress low-information channels.
IntroductionQuantification of [11C]CURB, an irreversible positron emission tomography (PET) radiopharmaceutical used to image fatty acid amide hydrolase (FAAH), requires invasive arterial sampling. Developing a non-invasive approach for [11C]CURB is challenging, since traditional reference region models are not directly applicable and the ubiquitous brain expression of FAAH complicates the identification of a ligand-free reference region. This study aimed to introduce and validate the irreversible reference tissue model (IRTM) used in conjunction with a data-driven, clustering-based white matter (WM) pseudo-reference region.MethodsIRTM was implemented using a coupled-fit approach to estimate a common k2′, resulting in two independent parameters per volume-of-interest: R1 and kf. Primary quantification was performed using the macroparameters Ri=Ki/K1′=R1k3/kf (relative net influx) and Rk3=λk3/K1′=R1k3/k2 (relative trapping index). Simulations characterized the sensitivity of IRTM to differences in cerebral blood volume between target and reference tissues. The model was validated against the gold standard arterial input function (AIF)-based quantification using retrospective [11C]CURB PET data from 10 healthy participants. A partition-based clustering algorithm was used to extract the centermost WM time-activity curve as a pseudo-reference region.ResultsIn the IRTM validation study with human data, Ri emerged as a more robust metric than Rk3, with the latter exhibiting high between-subject variability. Although simulations identified systematic biases in IRTM estimates driven by blood volume asymmetries, IRTM-derived Ri estimates showed strong concordance with AIF-based Ri (R2 = 0.96), λk3 (R2 = 0.90), and Ki (R2 = 0.96). Mean errors in Ri ranged from −11.2 ± 9.9% in the frontal lobe to −3.1 ± 10.4% in the amygdala. Voxel-wise parametric maps demonstrated high image quality and signal-to-noise ratio at the individual subject level.ConclusionThis study provides a validated, fully non-invasive framework for [11C]CURB PET quantification. Combined with a partition-based clustering approach to select a data-driven WM pseudo-reference region, IRTM achieved excellent concordance with AIF-based measurements, supporting the feasibility of IRTM for non-invasive FAAH imaging.
BackgroundOpen-door policies in inpatient psychiatric care have been increasingly adopted to enhance patient autonomy and reduce coercive practices. However, systematic methods for evaluating patient experiences in these settings remain limited. This study applied Patient Journey Mapping in an acute psychiatric unit operating a structured, time-limited (6 hours/day) open-door policy to comprehensively analyze the hospitalization experience.MethodsAn exploratory sequential mixed-methods design was used. Qualitative data were first collected through focus groups with patients (n = 18), following GRAMMS guidelines, to identify critical touchpoints across the hospitalization journey. Quantitative data were then collected from patients (n = 32) using an adapted Likert-type ordinal scale to assess satisfaction with the identified touchpoints. Data were analyzed using content analysis methods.ResultsThe analysis delineated the main phases of hospitalization and identified key experiential touchpoints, including “Moments of Truth,” “Moments of Pain,” and “Wow Moments,” which substantially shaped patients’ perceptions of care quality. Open-door policies were perceived as supporting autonomy and reducing psychological distress. However, several challenges remained, particularly regarding the admission process, privacy during medication administration, and communication about structured activities.ConclusionPatient Journey Mapping provides a useful patient-centered approach for evaluating experiences in open-door psychiatric inpatient settings. The findings highlight opportunities to optimize service delivery, strengthen therapeutic environments, and support recovery-oriented practices in acute psychiatric care.
IntroductionAttention-Deficit/Hyperactivity Disorder (ADHD) is a condition characterized by persistent patterns of inattention and/or hyperactivity-impulsivity. This case highlights the potential benefit of integrating Immersive Virtual Reality (IVR) with cognitive-behavioral therapy (CBT) in the rehabilitation of a child with ADHD. It contributes to emerging evidence by showing how a combined approach may simultaneously target executive, attentional, and motor domains within a single intervention.Case presentationAn 8-year-old child with deficits in sustained and selective attention, impaired executive functioning (including planning and working memory), impulsivity, and difficulties in motor regulation, as revealed during baseline assessments, impacting daily functioning. The patient was diagnosed with combined-type ADHD and underwent a 12-week CBT intervention, followed by integrated IVR-CBT intervention targeting executive functions and self-control, once a week for 12 weeks. The intervention was conducted using the CAREN (Computer Assisted Rehabilitation Environment), an immersive virtual reality platform integrating multisensory input and interactive tasks to promote cognitive and motor engagement. Post-intervention assessments showed improvements in sustained and selective attention, planning, working memory, and balance. There was also an increase in involvement and a reduction in impulsivity.ConclusionThe findings support the hypothesis that immersive, embodied interventions targeting both executive and sensorimotor processes may represent a promising novelty adjunctive rehabilitation approach. Further studies are needed to evaluate efficacy, generalizability, and to confirm these findings in larger samples. This case report was prepared in accordance with the CARE Guidelines.
Cognitive inflexibility, a reduced capacity to shift mental set or update behavior when circumstances change, is a significant feature of obsessive-compulsive disorder (OCD). However, while cognitive inflexibility is superficially a coherent entity, it appears to engage a range of distinct cognitive processes. This raises the question of how far deficits in different tasks involve the same core dysfunction and how far the dysfunctions are unique but superficially have similar results. Here we detail the neural basis of OCD deficits across eight tasks that challenge different aspects of cognitive flexibility: set-shifting (Wisconsin Card Sorting Task, Intra/Extra-Dimensional Set Shift), feedback adaptation (Reversal Learning), interference control (Stroop Color and Word Test), response inhibition (Go/No-Go Task and Stop Signal Task), working memory updating (n-back Task), and value-based flexibility (Delay Discounting Task). Cognitive inflexibility in OCD appears linked to functional abnormalities in a largely shared set of core structures (anterior cingulate cortex, caudate nucleus, orbital frontal cortex, and prefrontal cortex) and less shared ‘peripheral’ structures (putamen, thalamus, parietal cortex), based on cross-task convergence of OCD-specific abnormalities. Even the shared core appears to engage multiple interconnected neural networks. The cortico-striato-thalamo-cortical network appears central, and the salience network and the default mode network also contribute with indirect, task-varying effects on peripheral parts of each. Peripheral areas are less consistently involved but appear to contribute to OCD nonetheless. Cognitive inflexibility in OCD appears to involve failure of interaction between multiple networks rather than dysfunction in any single system.
ObjectiveThis study aimed to use a three-level meta-analysis to examine the effects of exercise interventions on executive function (EF) in children and adolescents with autism spectrum disorder (ASD), and to explore the relationship between exercise dose and intervention effects through subgroup analyses.MethodsA systematic search was conducted in Web of Science, PubMed, Embase, and the Cochrane Library from database inception to November 17, 2025. All included studies were randomized controlled trials. A three-level random-effects model was applied to integrate multiple dependent effect sizes. Hedges’ g was used to calculate effect sizes, and negative values indicated improvements in EF following exercise interventions. Study quality was assessed using the PEDro scale, and publication bias was evaluated using Egger’s regression and the trim-and-fill method.ResultsSeventeen studies involving 626 participants were included. Exercise interventions produced a moderate and significant improvement in EF (g = –0.34, p < 0.0001) with low heterogeneity (Q = 50.42, p =0.268). Subgroup analyses showed that only intervention duration significantly moderated the effects, with programs lasting ≥10 weeks yielding the greatest benefits.ConclusionExercise interventions significantly improve EF in children and adolescents with ASD, with the largest effect sizes observed in interventions lasting ≥10 weeks. Further large-scale studies are needed to clarify dose–response relationships and inform optimal exercise prescriptions. This meta-analysis was designed and reported in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA 2020) guidelines. Systematic review registrationhttps://www.crd.york.ac.uk/PROSPERO/, identifier CRD420251236112
