What & So What - The Biweekly Entangled Health (Tech) Update

Kickoff letter, which covers longer timeframe to start with.

Feb 07, 2026

Greetings

Hello readers, subscribers and friends. I am trying this format, It’s in my personal private (and subjective, limited) capacity (it’s the weekend!), and not company sponsored.

Let me know if you like it. The idea is to have a regular (biweekly?) update on what’s happening in the zone of “entangled health” - on top of the podcasts. Maybe a longer monthly or quarterly “digest”. Let’s see where we go, Entangled Health is not even a year old. A permanent experiment. The sorting of the news is not according to “importance”, and likely for some flavors a bit “too much AI” in the beginning, it’s moving into Wearables and Sensors and Quantum and Biofield, please be not shocked away, give it a chance.

So, as a starter, the news here are covering a bit longer timeframe and thus contain a bit more content than strictly the last two weeks, so e.g. the quantum computing papers are NOT form the last two weeks. Also AI Regulation is not really new, and “bubble burst” claims are here since the beginning of the first token. And the CES was early January, but at this was quite important event, it’s still covered (greetings to my employer here), but not exhaustively, as I was not there in person (-;

While I am writing this, I am checking how to best use substack to add “complementary material” - if someone has a best practice to use the “archive” or add “premium content” or so, PLEASE let me know. Thanks!

So, enough preamble - do not forget to do your own thinking - let’s start.

Image Credit: your favorite prompt1 and some transformers in some data center.

What’s Actually Happening

Clinical AI moves from pilots to randomized evidence – and exposes workflow fractures
A January 2026 Stanford review notes that randomized trials of clinical AI show physicians using decision support alongside standard resources make measurably better treatment decisions than controls, but real-world deployment still struggles with integration into existing clinical workflows and data systems. The evidence base is shifting from retrospective AUC charts to prospective impact on therapy choices and safety endpoints. Translation: AI is increasingly defensible in tumor boards and heart failure clinics, not just in slide decks. The catch: implementation debt (fragmented EHRs, alert fatigue, liability questions) is now the main bottleneck, not model accuracy.

AI Imaging Tools Quietly Accumulate FDA Clearances
GE HealthCare has received FDA clearance for an AI-powered tumor burden analysis tool that automates whole‑body quantification on PSMA PET/CT and SPECT/CT studies. The software standardizes measurement of metastatic prostate cancer lesion load, turning what used to be a highly manual, variable task into a repeatable quantitative pipeline that can support therapy selection and response tracking. Implication: “AI radiologist” narratives are giving way to more realistic narrow tools that make trials and follow‑up more objective. The catch: performance is modality‑ and indication‑specific, requires well‑curated imaging workflows, and sits inside a tightening regulatory regime for AI‑enabled devices.

Regulators quietly widen the lane for autonomous-ish AI agents in care
The FDA’s January 2025 draft guidance on AI-enabled device software functions formalized a lifecycle approach (data lineage, bias analysis, change-control plans), and a January 2026 update to Clinical Decision Support guidance signaled enforcement discretion even when software outputs a single preventive, diagnostic, or treatment recommendation, as long as clinicians can independently review the basis. That effectively clears room for “co-pilot” agents to suggest orders, draft notes, or flag high-risk patients with less fear of instant device classification. The implication: hospital IT can now green-light more autonomous workflow agents (triage, prior auth, documentation) under a clearer risk framework. The limitation: truly learning, self-updating agents that change behavior post-deployment remain tightly constrained and will still need heavy post-market monitoring and predefined update plans.

AI Agents Start Ordering Your Labs, Not Just Writing Your Notes
Oracle Health has upgraded its clinical AI agent from passive ambient scribe to an active workflow assistant that drafts prescriptions, labs, imaging orders, referrals, and follow‑ups in real time during visits. The system uses semantic reasoning over conversation plus chart data and has reportedly saved clinicians more than 200,000 hours since launch, with order-entry automation expected to further reduce “click burden.” Translation: we are quietly crossing from decision support into semi-autonomous care orchestration. The catch: this remains a proprietary EHR‑embedded tool with human-in-the-loop oversight and unknown generalizability outside Oracle’s ecosystem or U.S. health systems already heavily invested in its stack.

Wearable sensors graduate from step counts to multi-sensor clinical observatories
A 2026 systematic review of 30 studies in JMIR mHealth reports that 67% used inertial sensors, many combined with other modalities, to continuously track gait, mobility, and falls risk in Parkinson’s disease, stroke, multiple sclerosis, and frailty, with machine-learning models (mostly random forests and deep learning) reaching AUCs up to 0.97 for fall-risk prediction in some cohorts. Continuous, real-world mobility streams are now credible endpoints for rehabilitation and aging research, not just wellness metrics. This pushes care toward proactive intervention for decline before hospitalization. The limitation: heterogeneous sensor placements, small sample sizes, and non-standardized analysis pipelines make cross-study generalization and regulatory acceptance tricky, and most work is still in single-center or short-term trials.

Quantum computing in drug discovery inches from promise to playbook, still hits hardware walls
A January 2026 Nature Review outlines how quantum approaches could slot into molecular simulation, drug–target interaction prediction, and clinical-trial optimization, arguing that quantum devices around the 100-qubit scale are beginning to support more realistic benchmarks when paired with classical and AI methods. The narrative is less “magic black box” and more tiled into specific steps: better protein–ligand energies, smarter library pruning, and more efficient trial designs. The implication: pharma pipelines will likely see quantum first as a hybrid acceleration layer in discovery and trial design, not as a standalone engine. The catch: current noisy devices cannot yet model large biomolecules with chemically accurate resolution, so genuine “quantum advantage” in therapeutics still depends on future error-corrected machines and careful problem selection.

Biofield and holistic practices edge into more rigorous stress and well-being studies – mostly around caregivers
A 2025 cluster randomized controlled trial in a level 1 trauma-center nursing cohort (128 RNs) found that Healing Touch sessions produced statistically significant improvements in nurses’ mid-shift stress scores versus a deep-breathing control, measured via vital signs and a visual analog stress scale across repeated time points.

This anchors at least one biofield-linked modality in a reasonably powered, real-world trial focused on caregiver resilience rather than metaphysical claims. The implication: health systems facing burnout may quietly adopt such interventions under the banner of “holistic support,” even while mechanisms remain speculative. The limitation: outcomes are short-term stress measures in a specific population, not patient-level clinical endpoints, and the broader biofield measurement landscape is still early-stage, fragmented, and underfunded, with most devices (biophoton, gas-discharge visualization, acupuncture-conductivity tools) lacking large validation studies.

Biofield Tech Is Getting an AI Upgrade, But Not Yet Clinical Validation
Vendors in the biofield space are now claiming that machine‑learning models can detect individual “biofield fingerprints” from biophoton and electromagnetic measurements with up to 94% accuracy and predict health imbalances months before symptoms. This builds on earlier biophysics work suggesting ultra‑weak electromagnetic and photon emissions from living systems and on a cottage industry of biofield imaging and sensing devices. Implication: there is growing interest in treating subtle energy as a measurable signal stream that could be correlated with meditation, emotional states, or healing practices. The catch: current results are largely outside mainstream peer‑reviewed clinical research, with small or unpublished datasets, unclear reference standards, and no major regulator recognizing biofield metrics as clinical endpoints

Digital twins stay mostly conceptual for individuals, but the stack is forming underneath
While full patient-specific digital twins remain more roadmap than routine, the combination of high-frequency wearable data, AI-enabled clinical decision support, and emerging EU AI Act implementation is laying the infrastructure: continuous streams of physiological signals, standardized risk models, and a regulatory frame for high‑risk predictive systems. Industry analyses in late 2025–early 2026 emphasize integrating sensor, imaging, and genomic data into “virtual patient” models for trial simulation and therapy optimization, with quantum and AI methods proposed for handling the combinatorial explosion in parameter space. The implication: expect digital twins to appear first as modular, organ- or pathway-specific simulation services used by pharma and specialty centers, not as full-body avatars in your patient portal. The limitation: data harmonization, model interpretability, and validation across diverse populations are unresolved, and regulatory pathways for using such twins to guide real treatment decisions are still being defined.

The Players Moving Fast

FDA & EU Commission: Redrawing the AI guardrails

The tech: The FDA’s AI-enabled device guidance and updated CDS rules embed lifecycle management, bias analysis, and change-control into submissions, while the EU AI Act classifies most clinical AI as “high-risk,” demanding robust risk-mitigation, transparency, and human oversight.
Why it matters: Together, they de-risk investment in clinical AI, AI agents, and future digital-twin systems by clarifying what “acceptable” looks like in the US and EU, especially around continuously learning models.
The timeline: AI Act provisions phase in through 2026–2027, with high-risk system obligations and GPAI codes of practice taking effect within 12–24 months; FDA’s guidance is already shaping 2025–2026 submissions.
What’s missing: Harmonized global standards, practical templates for multi-site post-market monitoring, and clear lines around responsibility when autonomous agents make recommendations that clinicians routinely accept.

Academic–hospital consortia in mobility and chronic care: Turning wearables into clinical endpoints

The tech: Multi-sensor platforms (inertial units, smartwatches, sometimes microneedles) streaming gait, tremor, and cardiometabolic data into machine-learning models to predict falls, disease progression, or rehab response, with some studies reporting AUCs near 0.97 for specific tasks.
Why it matters: This is the practical substrate for digital twins, AI agents, and remote-first care: dense, longitudinal, context-rich data rather than episodic vitals. It also reframes mobility and function as measurable, reimbursable outcomes.
The timeline: More robust, multi-center trials and payer interest seem plausible over the next 2–4 years, especially for Parkinson’s, frailty, heart failure, and post-stroke monitoring, with first regulated solutions already on the market and being refined.
What’s missing: Standardized sensor placements and protocols, reproducible ML pipelines, integration into EHRs and clinician workflows, and convincing evidence that acting on these signals changes hard outcomes (hospitalizations, mortality) rather than just generating alerts.

Quantum–pharma collaborations : Building the hybrid pipeline

The tech: Hybrid quantum–classical workflows for molecular simulation and lead optimization, where quantum routines approximate electronic structure for selected subsystems while classical/AI models handle large-scale screening and ADMET prediction.
Why it matters: If they can reliably shave months off hit-to-lead and reduce late-stage attrition, the economic incentive is enormous for oncology, CNS, and rare disease pipelines where failure is expensive.
The timeline: Expect early “quantum-assisted” case studies (e.g., a few targets per portfolio) over the next 3–5 years, with broader impact contingent on more stable hardware and tighter integration into existing cheminformatics stacks.
What’s missing: Demonstrations of clear, task-specific quantum advantage on therapeutically relevant systems, transparent benchmarks against best-in-class classical methods, and workflows robust enough for regulated environments rather than research prototypes.

Oracle Health: Turning AI scribes into workflow agents

The tech: An EHR‑embedded clinical AI agent that uses ambient listening plus chart data to generate structured documentation and now draft orders (meds, labs, imaging, referrals, follow‑ups) in real time during encounters.
Why it matters: Moves from “note‑taking assistant” to semi‑autonomous workflow execution, directly attacking clinician burnout and throughput constraints rather than just documentation convenience.
The timeline: Already deployed in U.S. sites using Oracle Health, with incremental capability rollouts likely over the next 12–24 months as health systems validate safety and ROI in specific specialties.
What’s missing: Public, peer‑reviewed evidence on error rates, override patterns, and patient‑safety outcomes, plus clarity on liability when agents suggest or mis‑draft orders in edge cases.

Roche Diagnostics: Industrializing digital twins

What they did: Published a strategic framing of digital twins for healthcare, emphasizing virtual replicas that ingest imaging, lab, sensor, and genomic data to model disease progression and simulate treatment response, particularly in complex conditions.
The tech: Specialist algorithms over multi‑modal data streams, creating a dynamic feedback loop between patient and model and enabling scenario testing (“virtual trials”) at the individual level.
Why it matters: Offers a pragmatic bridge from population‑level risk scores to individualized simulation, with obvious applications in trial enrichment, therapy selection, and surgical planning.
The timeline: Expect meaningful adoption first in research and high‑value tertiary centers over the next 3–5 years, with broader clinical use gated by validation standards and integration into guidelines and reimbursement.
What’s missing: Robust prospective evidence that twin‑guided decisions outperform standard care across diverse populations, standardized data models, and clear regulatory categories for when a twin becomes a regulated medical device.

CES 2026 convergence panels (Laura Matz, Merck KGaA & partners): From gadgets to quantum‑ready health stacks

What they did: They used CES 2026 to stage “convergence in action” panels where Laura Matz and partners walked through concrete use cases linking AI, semiconductor innovation, cloud data platforms, and quantum‑ready compute to re‑engineer drug discovery and the lab of the future.
The tech: Cross‑sector “convergence in action” where AI, advanced data platforms, semiconductor innovation, and quantum‑ready compute are woven together with biology and chemistry, showcased through panels on disrupted drug discovery and the “lab of the future.”
The focus: using shared data backbones and automation to link materials, electronics, and wet‑lab workflows so predictive models can guide experiments rather than merely document them.
Why it matters: If this stack actually matures, it shifts CES from a consumer‑gadget show into a neutral ground where chip makers, cloud providers, and pharma can prototype the infrastructure that underpins quantum‑assisted discovery, complex‑disease pipelines, and high‑throughput translational research. The economic and clinical upside is obvious: fewer blind alleys, faster iteration on tough indications, and better reuse of experimental data across sectors.
The timeline: Over the next 3–5 years, expect more “lighthouse” collaborations, i.e. single therapy areas or platforms where AI‑driven and quantum‑adjacent workflows are tightly integrated with automated labs before anything resembling a standard, plug‑and‑play stack emerges for mainstream pharma and health systems. Real impact will likely show up first in R&D and trial design, long before frontline clinical workflows notice.
What’s missing: Hard, transparent evidence that these convergent stacks consistently shorten development timelines or reduce attrition compared with best‑in‑class classical and AI‑only approaches, plus clear governance for how quantum‑ready infrastructure, proprietary data platforms, and regulated clinical pipelines will coexist without locking innovators (and patients) into a few vertically integrated ecosystems.

Biofield/Consciousness Tech Startups (Bio‑Well & peers): Instrumenting subtle energy

What they did: Positioned AI‑enhanced biofield imaging and sensing platforms as tools to capture individual “energy signatures,” track changes during meditation or healing, and correlate patterns with emotional and health states.
The tech: Optical and electromagnetic sensing of ultra‑weak emissions (for example, biophotons) combined with pattern recognition algorithms to cluster or classify “field” configurations.
Why it matters: For consciousness and holistic health communities, this represents an attempt to move from purely experiential claims to repeatable measurement—potentially bridging meditation, energy medicine, and conventional physiology.
The timeline: Near‑term impact is in wellness, coaching, and research settings, with any serious clinical role contingent on rigorous trials tying biofield metrics to hard outcomes over the next 5–10 years.
What’s missing: Standardized protocols, independent replication, large‑scale clinical datasets, and any recognition of these signals as valid biomarkers by regulators or major medical societies.

So What Does This Mean?

We are watching three layers of the stack harden at once: sensing at the edge (wearables, biofield-adjacent devices), AI and agents in the middle, and quantum plus simulation at the high-compute core.

Continuous monitoring is no longer science fiction. The main questions now are what to measure, how to standardize it, and whether acting on it improves real-world outcomes rather than simply generating more dashboards. At the same time, regulators are moving from “wait, what is this?” to detailed checklists, which paradoxically both constrain and accelerate innovation by rewarding players who can document bias, drift, and lifecycle plans as first-class engineering tasks.

For consciousness researchers and biofield enthusiasts, the pattern is more nuanced. The best-designed studies are gravitating toward stress, burnout, and functional outcomes in caregivers or chronic-illness populations, where subjective experience is central and small improvements matter operationally. Measurement work is still dominated by electromagnetic, biophotonic, and electrophotonic approaches rather than “subtle energy” per se, and most devices remain early-stage with limited replication.

The opportunity is to plug these modalities into the same rigorous data and regulatory scaffolding as wearables and digital therapeutics. Standardized protocols, pre-registered endpoints, and explicit hypotheses rather than trying to live outside the clinical evidence ecosystem.

Quantum and digital twins sit more in the “invisible infrastructure” layer: they will shape how therapies are discovered, tested, and priced long before patients consciously interact with them.

You know I am skeptic here myself and work on investigating different “quantum” and “hybrid” methods on the theoretical (Mathematics, oh-oh!) and practical (Hardware engineering and affordability to execute) side. So, Early quantum-drug discovery work suggests real potential to reshape preclinical search space, but hardware realities keep timelines conservative, with most near-term gains coming from hybrid schemes and clever problem encoding. We still lack algorithms.

Digital-twin rhetoric in marketing decks often outpaces the current reality, yet the components - i.e. high-fidelity longitudinal data, regulatory clarity for high-risk AI, and interest from payers in virtual trials - are starting to lock into place. Clinically, expect “partial twins”: heart-failure trajectory models, neuro-degeneration progression simulators, or therapy-response predictors in oncology, each tightly scoped and heavily audited.

The most material short-term shift for health systems is not a single breakthrough but a reconfiguration of who (or what) does the cognitive labor. AI agents drafting notes, prioritizing in-baskets, and nudging clinicians about gaps in care are easier to deploy under the new FDA and EU AI regimes and are already being piloted across hospitals. These tools don’t require new physics; they require trust, monitoring, and alignment with reimbursement and staffing realities. The upshot: in the next 2–3 years, your “clinical quantum leap” is more likely to be a quietly competent AI co-pilot connected to your EHR than a room-temperature quantum computer diagnosing cancer.

The Irony Check

We are spending billions to simulate entire patients as digital twins and optimize drug pipelines with quantum-enhanced algorithms, while the most robust biofield result on the board right now is that a hands-on Healing Touch session can reliably lower nurses’ stress scores during brutal shifts.

Regulatory guidance for AI agents runs to hundreds of pages specifying bias quantification, post-market drift, and change-control plans, yet the agents hospitals actually rush to deploy are the ones that schedule appointments and clean up documentation so humans can get through their day.

And as we edge toward quantum-enabled precision medicine, the data that often makes or breaks a real clinical decision is still a very classical time series from a wrist sensor showing how far, and how steadily, someone actually moved this week.

Walk the Talk

Which development caught your attention?

Reply with the topic you’d like unpacked in a podcast episode (no promise, though).

Topics people are asking about:

Quantum-assisted drug discovery vs. classical AI in real pipelines
From wearables to digital twins: when does continuous sensing actually change outcomes?
Biofield therapies at the bedside: what we really know about mechanism and clinical impact so far
Or suggest your own (e.g., AI workflow agents under the new FDA and EU AI rules).

Hit reply. Let’s talk.

Thomas Ehmer

Discussion about this post

Ready for more?