iiinova — where your digital self keeps evolving

"De toekomst behoort toe aan probleemoplossers" ... en morgen komen die probleemoplossers samen in Antwerpen voor #careandcode www.uptodatewebdesign.com/2026/09/zor...

000

Tavus Griffin: a ‘Human Interaction Model’ that renders a lifelike person who can hear, see, talk, and react over live video. This is getting spooky good, very fast. www.tavus.io/griffin

000

If you keep the dead happy, they are less likely to haunt you. liquiddeath.com/en-be/pages...

000

AI shifts the founding bottleneck from execution to selection. Steve Blank's point lands hard: AI makes bad ideas go faster. The moat is now judgement, not speed. abundance entrepreneurship customer discovery startup da.gd/ezHV

The Handover #9: the AI search that misses the paper. University of Florida researchers tested five AI evidence-search tools: 12% of the relevant evidence was never retrieved by any of them. cliniciansthatcode.com/briefing/is...

Bart Collet @atbart.bsky.social · 01/10/2026

1.7 billion pounds sits stored on Starbucks cards and apps, voluntarily, before anyone's ordered a coffee. That's design, not loyalty. Preventative healthcare needs the same: a health menu, points for healthy behaviour. preventative health tinyurl.com/285goghl

000

10,000 AI instances coordinated to solve what no single model could. Collective AI behaviour is the real risk surface; not smarter models, but emergent swarm dynamics. No intent required.

000

Bart Collet @atbart.bsky.social · 01/10/2026

The Doorman Fallacy: valuing a role only by its replaceable tasks. A doorman does more than open doors. A doctor does more than diagnose. AI should strip tasks that never needed a human, not replace the ones that do. AI in healthcare tinyurl.com/2ym583t7

000

Bart Collet @atbart.bsky.social · 01/10/2026

Neko Health opened in New York with 25,000 people on the waitlist. Waiting list in traditional healthcare: broken system. Waiting list at a startup: proof of demand. Same queue. Different story. da.gd/Enp3y

000

Bart Collet @atbart.bsky.social · 30/09/2026

Management still thinks a Claude or OpenAI subscription for everyone will magically transform the company. PwC (50,000 workers): 80% of front-runners get AI training. In the 56% engine room, only 40% do. A licence is not a capability. da.gd/D4UYO

000

Bart Collet @atbart.bsky.social · 30/09/2026

30 sources in and the finding is structural: small local models pay back fast, large hospital GPU clusters often lose to cheap APIs. Tiered proximity AI architecture may be the only honest answer. hospital AI inference cost build vs buy open models da.gd/It3g3

000

Bart Collet @atbart.bsky.social · 30/09/2026

"Let's explore synergies" is not a partnership. No joint product, no named decision-maker: that's how cross-industry deals quietly stall. The fix: one-line joint product, someone who'd take the call. corporate innovation tinyurl.com/253bynp7

000

100 is the new 80. Bernard Gilbert, 100, led a 200-strong flash mob through Worcester for an Alzheimer's charity. His secret: 90+ years of dancing. Start early. Keep moving. uk.news.yahoo.com/britains-ol...

Finally, AI is being deployed in hospitals. And the first real ROI: ~$942M in extra costs (BCBS claims, 2023-25). More diagnoses on paper, no change in care. Insurer's analysis, so salt to taste. www.bcbs.com/about-us/as...

000

McDonald's and Starbucks retraining staff to restore human service; after building the tech to remove it. Efficiency worked. Belonging didn't follow. Fast food customer experience automation human touch retail behaviour

Bart Collet @atbart.bsky.social · 28/09/2026

47 longevity events already locked in for 2027, first one 15 January at Stanford. 2026 still has 134 to go, October holding 68. Circuit now: 550+ events, 40+ countries. longevity events tinyurl.com/2cs8qknm

000

Naive AI use in innovation pipelines deepens bottlenecks rather than removes them across ideation, screening, consumer insight, and market learning. Faster confirmation of a prior is not the same as learning. da.gd/WXd2TZ

000

Bart Collet @atbart.bsky.social · 28/09/2026

Your phone can't get online. Your car runs on steam. Absurd? That's healthcare: a reactive 'Sickcare OS' while every other industry digitised. A real fix needs new entrants building from scratch, not more panels. healthcare innovation tinyurl.com/27gl83mu

000

Bart Collet @atbart.bsky.social · 27/09/2026

A16z launching a one-year tech academy as a university alternative. The AI skills argument is solid; the VC-as-educator incentive structure is the part worth scrutinising. Talent pipeline and education reform in the same vehicle.

000

Bart Collet @atbart.bsky.social · 26/09/2026

Joining Hackathon Hub as Ambassador for HealthTech. Healthtech hackathons across Europe rarely talk to each other, something I see through Care & Code. Hackathon Hub is fixing that. hackathonhub.eu

000

Bart Collet @atbart.bsky.social · 26/09/2026

3,199 miles, zero control inputs. Joby's autonomous cross-country flight is notable less for the distance than for the use cases named: freight, medical logistics, disaster response. Cargo autonomy is the proving ground, not urban air mobility. da.gd/zcXMT

000

Joby just flew the first fully autonomous eVTOL flight across the US. Can I book this via Uber, Lyft, Gett, Bolt... or is it its own silo app forever? www.jobyaviation.com/news/joby-c...

000

Retailers quietly rehiring cashiers after self-checkout expansion; shrink and scanning errors drove the reversal more than customer friction. Classic technology adoption arc: savings projection, underestimated operational cost, retreat. AI enterprise rollouts are mid-loop on the same curve.

000

Healthcare's labour shortage is a demographics problem, not a hiring problem. AI scribes treat the symptom. The uncomfortable fixes sit at the fertility-tech end: fewer babies now means fewer clinicians in 25 years. da.gd/gFn6Z

000

The Handover #8: what a chatbot owes a patient in crisis. Mass General and UCLA publish the safety router behind an opioid-recovery chatbot: three risk tiers with a human handoff built in. cliniciansthatcode.com/briefing/is...

000

Function Health's biomarker data now feeds Meta's Muse glasses. Biomarker testing + a personal AI agent watching your numbers. Where will the new blue zones get built? Postal code might matter more than your genes. longevity.technology/news/functi...

Virtual CAIO that interviews employees and maps cost centres before deploying automations. Runs on your own servers. Early customers reporting millions in added profit ahead of schedule. Enterprise AI transformation as operating system, not consulting engagement. codos.ai/

Roughly the same number of longevity events in Europe and the US. Nothing else matches: Europe debates how ageing works, the US debates what to ship today. 500+ events tracked across 40+ countries. da.gd/2gJdp

Bart de Witte (Isaree): zorgverleners bezitten hun AI beter dan ze te huren. Bij een eigen, open model kost de 200e taak evenveel als de eerste: niets. Op 3 okt op Clinical Build Day: careandcode.be/blog/own-yo...

000

"Gradually, then suddenly." (Hemingway) Oura started as a sleep ring. Now it's wiring wearable data into multi-omic diagnostics and physician-guided care. Not a feature. A category change. longevity.technology/news/oura-a...

000

Napster rebranded as an AI education platform, building digital teacher agents for Gulf region schools. The brand logic: notoriety is cheaper than building trust from zero. Unclear if that converts in edtech the way it might in consumer tech.

000

Which AI infra signs a HIPAA BAA with no upgrade needed? Firebase/GCP and the big three model APIs (Azure OpenAI, Bedrock, Vertex). Vercel, Supabase, OpenAI, Anthropic: a paid tier. Netlify, Mistral, Clerk, Resend: Enterprise only. cliniciansthatcode.com/guides/whic...

McDonald's x Krispy Kreme: $28.9M in charges, dead within a year. Nintendo x Sony died on stage and created the PlayStation. Unusual partnerships die of the boring checks nobody ran. da.gd/hdKxNT

000

AI doesn't kill people? Think again. Bloomberg found the Pentagon's own probe blamed flawed intel and an overreliance on AI for the strike that killed 123 children in Minab, Iran. da.gd/zNuP1

000

AI ROI becomes measurable when you start with business goals, not AI goals. Break goals into KPIs, identify the work that drives them, then find where AI accelerates that work. Most organisations skip steps one through three entirely. da.gd/SeVH

000

Dating apps now screen for fitness. ATEAM is invite-only, picked by a committee of trainers and tastemakers. Same question I ask about longevity: who ends up outside the club? www.dazeddigital.com/beauty/arti...

000

Apple's health chief reportedly wears an Oura and a Whoop. The screenless Watch Air rumor reads like a confession: the Series 12 Readiness app is already a Whoop, minus the $199 a year. da.gd/yCg02

000

a16z built a 20,000-GPU cluster to win AI deals. The talent followed the compute. IKEA hires inside Roblox; Mayo Clinic's 50+ AI patents pull clinicians in. Nobody competes on job postings any more. da.gd/FJpHjF

Apple just bought assets from Sonera, a Berkeley spin-out with chip-sized sensors that read muscle and nerve signals without skin contact. Which Apple device gets it first? Or a new one? www.mobihealthnews.com/news/apple-...

000

Alibaba open-sourced Damo Radar: it reads abdominal CT scans and flags 146 conditions, cancers included. AUC 0.913 on ~40,000 real exams, ahead of 23 of 26 radiologists. Published in Science. da.gd/sb8UZX

000

A product management definition from ten years ago still circulates unchanged. AI compresses prototype cycles but doesn't touch the core job: deciding what to build and for whom. Speed was never the hard part. da.gd/XFCd2a

000

Bart Collet @atbart.bsky.social · 20/09/2026

Eroom's law: drug discovery gets slower and more expensive every year. Q1 2026 says AI might break it: 173 AI-discovered programmes in trials, 80-90% Phase I success vs 40-65% historically. da.gd/TlRTq

000

Software moats: replication cost, switching friction, network effects. All three weaken as AI lowers engineering cost. Enterprise software may barbell like 1990s newspapers; dominant platforms per buying centre, tiny niche tools, almost nothing viable in between. da.gd/SBoYyq