Skilled nursing as a stress test for clinical AI is a sharper framing than most. Longitudinal resident records give real outcome signal; but staff are too stretched to tolerate tools that add friction before reducing it. The "chase value, not AI" principle is either a useful filter or a new cliché.

Our longevity events database held 340+ events when we wrote up the methodology in February. Today: 450+ events, 3,700+ speakers, 40+ countries. A hundred new longevity events in five months. The circuit is growing faster than most insiders realise.

Fintech rebuilt money. Streaming rebuilt entertainment. The shopping trolley is still dumb, and it's the most health-consequential interface we touch daily. Food tech is a $184bn market; making the healthy choice the default is still unsolved.

Stanley x LoveShackFancy sold out in hours. Nike x SKIMS became a joint brand. Mastercard pulled 85 crypto firms into one ecosystem. The best B2B partnerships look absurd on paper; the audience overlap is the tell, not the product category.

Apple's on-device model activates 1-4B parameters out of 20B. PrismML ran a dense 27B model on the same iPhone with all parameters active. The assumed thermal and memory hard limits on mobile AI may be softer than the industry priced in.

HIPAA risk in AI coding tools is a prompt hygiene problem, not a product licensing one. Cursor packages code, file paths, and workspace metadata into every request. The compliance line sits at what your team pastes, not what tier they pay for.

Healthcare orgs are sitting on an under-used asset: video. Live-streamed surgeries, AI-personalised prep videos, behind-the-scenes trust-building footage. Same format can do patient education, staff training, and branding at once. Most orgs only use it for one.

Thrivetopia: a theme park where pedalling powers your rollercoaster and a meditation chamber gives you live biofeedback. The bet: preventive health education fails not from lack of content but lack of a format people want to show up for.

Spa sector adopting longevity framing: regeneration, contrast bathing, physiological resilience. The research base for thermal contrast is real. The question is whether premium hospitality is operationalising the science or borrowing the vocabulary. Hard to tell from the outside.

GenAI isn't like the printing press, or electrification, or the PC. It's like all three at once: lowering barriers to creation, reorganising how work gets structured, and spawning a new tool ecosystem, simultaneously. That's the part most takes miss.

Interesting reframe: hospital governance committees aren't blockers for clinician developers; they're what converts shadow AI on patient data into something sanctioned and scalable. The building isn't the risk. The unsanctioned building is.

Samsung's Galaxy Watch is adding Vascular Load and Antioxidant Index monitoring ; prevention metrics, not fitness metrics. Wearables are quietly shifting from performance tracking to early detection. Connecting that insight to care pathways remains unsolved.

Category-defining partnerships rarely come from the same industry. Stanley (tumblers) x LoveShackFancy (fashion) sold out in hours. Nike couldn't win a female audience in 50 years; SKIMS could, so NikeSKIMS exists. The further apart two firms sit, the higher the energy.

Healthcare's AI reflex: appoint a Chief AI Officer, call it strategy. But nobody has a manual for this yet; a title can't create missing knowledge. The real expertise is the 1-2% already building tools, like the ICU nurse with a GPT macro. That's your next pilot.

Most B2B partnerships don't fail at partner selection. They die inside: sales still on individual quota, the roadmap locked 18 months, the partner routed to someone who never heard of the deal. Internal readiness, not partner fit, is the real bottleneck.

Neko measuring visceral fat without X-ray or DEXA is the real claim to stress-test. Combining that with wearable data longitudinally is compelling for preventive health; it also creates a dataset that goes well beyond what most users will think about at sign-up.

Genome sequencing: €95M in 2001, ~€200 today. A 99.99% collapse. The €25k/year concierge healthcare bundle rides those same curves as solar and LED. The breakthroughs already happened; we're waiting for distribution, not science.

Never too hot for a healthcare hackathon #welldata

Dubai created a Longevity Authority. Saudi Arabia spending $1B/year on healthspan. Yet 15 of 430+ longevity events sit in the Middle East. Under 3%. Capital arrived. Calendar hasn't.

A startup raised $1M turning live sperm racing into a betting spectator sport. Absurd, until you note the Drone Racing League sold for $250M and US research grants face 40%+ cuts. Attention is a funding mechanism we keep underrating in health innovation.

430+ longevity events, 60+ countries, and far more speaking slots than speakers. A small recurring core runs the flagship summits; the rest appear once. The circuit has become an industry. https://dub.sh/0Z3d0Kq

Model triage is quietly becoming a structural advantage. Frontier models like Claude Fable for planning and review; cheaper models for execution. Loop engineering is the term emerging for this. The teams who get the routing right will have unit economics others cannot easily replicate.

GLP-1 drugs were built to fight obesity. Their rise is boosting fast food stocks. The wellness industry runs on these flywheels of unhealthiness: products that feed the very problem they're sold to solve.

Time to break some things

Kodak built the first digital camera in 1975 and buried it. Blockbuster passed on Netflix for $50M. The pattern is repeating with generative AI, just faster. The hard part was never the breakthrough; it's cannibalising your own cash cow.

Sakana Marlin runs up to 8hrs autonomously; hypothesis formation, verification, causal mapping, no steering. Beta was 300 finance/consulting folks. Does it produce compressed strategy insight or just a long doc that still needs a sharp human?

Nine big tech companies launched healthcare AI in 72 days. None claims to diagnose. All admin: lab results, scheduling, documentation. The barrier to clinical AI isn't technology; it's liability.

Dexcom buying Nutrisense pivots biosensing to behaviour change as a subscription. AI coaching ~£300/yr vs dietitians ~£4,000/yr. Unit economics are clear; clinical liability when they disagree is not.

Clinical AI tools built for doctors are being outscored by general-purpose LLMs on medical benchmarks. GPT-5 topped a 1,000-item evaluation over purpose-built tools like OpenEvidence. The real issue: no mandatory independent evaluation before deployment in patient care.

AI coding tools invent package names that do not exist. Attackers register those names and wait. Veracode found vulnerabilities in 45% of AI-generated code. The threat model for vibe coding is genuinely different from pre-AI security thinking.