←
Bart Collet53,156 posts
Bart Collet avatar

Bart Collet

@bart

‾‾‾|‾‾‾-shaped Innovator | Health & Tech | 🏃🏻 Workshops, hackathons, rapid prototyping 🚀 powered by generative AI & automation

◷ Joined June 2008▧ 5,871 Photos and videos
Bart Collet avatar
Bart Collet@bart·Oct 7

18 platforms checked against their own terms. None of the AI app-builders sign a BAA. Lovable, Bolt.new, Base44, Replit and v0 all sit at not-applicable: a BAA is a property of the infrastructure holding the data, not the tool writing the code. The layer underneath signs. Firebase and the big three cloud model APIs by default; Vercel, Supabase, OpenAI and Anthropic on a paid tier; Netlify, Mistral, Clerk and Resend at Enterprise only. No BAA means no PHI on that vendor, on any plan, at any price. Every row carries a verified date and a watcher that re-checks for drift. Which row surprised you? cliniciansthatcode.com/guide…

Post attachment
Bart Collet avatar
Bart Collet@bart·41m

ZAS (Belgium's largest hospital) and Coforix put an AI in the EHR that reads consultation letters and proposes SNOMED-coded diagnoses for the problem list. 9 out of 10 coded correctly. ~10 seconds per letter. Live in 11 specialties. The doctor still decides what goes in. Unglamorous plumbing, but it's what EHDS actually runs on. Over two thirds of Belgian hospitals don't maintain a coded problem list yet. nieuws.zas.be/ai-helpt-artse…

Bart Collet avatar
Bart Collet@bart·7h

The EU's MDR reform makes medical software Class I by default. Read the proposed wording closely and software that informs a clinical decision still mostly lands in Class IIa. That's one finding in a new npj Digital Medicine comment (Freyer, Gilbert et al.): well-targeted on paperwork and double regulation, but the AI-specific problems stay open. Change control plans only cover updates you can predict at submission. A model that keeps learning, or an agent querying PubMed, can't be pre-specified. So do we certify the device, or the process that keeps changing it? nature.com/articles/s41746-0…

Bart Collet avatar
Bart Collet@bart·7h

Mayo Clinic has just handed its CEO role to a radiologist who has spent 20+ years across medicine, research, strategy and AI. The pairing is telling. Radiology is where pattern recognition meets early detection, and Dr Callstrom's stated aim for AI is to catch disease early enough to intervene, not to automate paperwork. That is a more demanding bar than most health systems set. It is measurable, and it can fail visibly. Will early-detection AI hold up once it leaves a research hospital and meets ordinary clinical workflows? info.himss.org/dc/NDIwLVlOQS…

Bart Collet avatar
Bart Collet@bart·11h

None of the AI app-builders sign a BAA. Lovable, Bolt.new, Base44, Replit and v0 all sit at no; a BAA belongs to the infrastructure, not the tool generating the code. 24 platforms checked against their own terms. Firebase/Google Cloud and the big three cloud model APIs sign for any paying account; Vercel, Supabase, OpenAI and Anthropic on a paid tier; Cursor, Claude Code and four more at Enterprise only. New last week: voice. Twilio and Vapi sign on a paid tier, ElevenLabs and Deepgram at Enterprise only. Where does your stack sign? cliniciansthatcode.com/guide…

Bart Collet avatar
⟳ Bart Collet retweeted
Bart Collet@bart·Nov 13

The Handover #14: who checks that AI devices work. The FDA has cleared 1,614 AI-enabled devices; almost none have data on patient outcomes. Also: a Kenyan clinical officer's claims agent and a spine surgeon's DICOM viewer. cliniciansthatcode.com/brief…

Post attachment
Bart Collet avatar
Bart Collet@bart·19h

The watch can see you now. Cornell's EchoMotion turns an off-the-shelf smartwatch into sonar: the built-in speaker and mic bounce ultrasound off your body, and a lightweight AI model reconstructs your full-body pose. No cameras, no extra hardware. Use cases: workouts, daily activity, rehab progress. Next up: pairing it with smart-glasses vision to track the objects you handle. news.cornell.edu/stories/202…

Wouter Verschelden avatar
⟳ Bart Collet retweeted
Wouter Verschelden@woverschelden·Nov 17

🥊 Hoe Vandenbroucke in één interview alles verder stuk slaat: zelfs Prévot kreeg een dolk in de rug, Arizona op sterven na dood 🗡️ Vooruit zet zich nog meer in een één-tegen-allenpositie: hoe valt dit nog te lijmen? 🎙️ Goedemiddag. Een tv-interview met Vooruit-vicepremier Frank Vandenbroucke zindert na in de Wetstraat. Vooruit trok gisteren naar de tv-studio’s, om zich te verdedigen na een mislukking van Arizona dit weekend, waarbij de Vlaamse socialisten geïsoleerd stonden, vier tegen één. Maar wat een signaal moest worden om aan te geven dat ze niet met de PS in zee willen, draaide uit op iets heel anders. ⚔️ Vandenbroucke viel niet alleen premier Bart De Wever (N-VA) openlijk aan (“het was te nemen of te laten”), hij haalde ook Georges-Louis Bouchez (MR) onderuit (“een oubollig belgicistische Franssprekende man die nooit Nederlands wil spreken”), maar de venijnigste aanval was gericht tegen Maxime Prévot (LE), die via de index “Vlaamse leerkrachten laat opdraaien voor de slechte begrotingen in Franstalig België”. 🚌 “Zijn grootste vriend aan tafel smijt hij zonder verpinken nu onder de bus?”, zo keek men binnen Arizona vol ongeloof naar wat er te zien was. “De leugens die hij nu durft te komen vertellen”, zo klinkt het hard over Vandenbroucke, die aan de onderhandelingstafel al langer een flinke reputatie opbouwde van “toxisch gedrag”. 🧭 Bij de N-VA keek men dan weer met grote ogen naar de hernieuwde “Vlaams-nationale reflex”, die Vandenbroucke zich aanmat in het interview. “Sinterfrank wil niet aan het kunstenaarsstatuut raken, dat voor 75 procent vol zit met Franstaligen.” 🚪Overigens leidde het openen van de deur naar de N-VA bij de PS, om zo Vooruit gerust te stellen, al tot een flinke clash intern, tussen Paul Magnette (PS) en Ahmed Laaouej (PS). 🧊 Bij Vooruit zelf klinkt het dat Vandenbroucke ook “diep gekwetst is door de houding van de anderen”, als verklaring voor de uitvallen. Maar dat de schade er andermaal is, ontkent niemand. Zowel Les Engagés als cd&v hopen ondertussen dat “de redelijkheid weerkeert”. “Maar hoe lang kan je nog zeggen dat we aan de afgrond staan, we zitten er al lang over”, zo zucht men binnen de regering. ⚖️ Het parket-generaal in Brussel opent een onderzoek naar de persoonlijke rol van minister van Justitie Annelies Verlinden (cd&v) in de zaak rond i-Police. Dat is vrij ongezien: de ‘eigen’ diensten die een minister onderzoeken. Volgens Verlinden is er geen vuiltje aan de lucht, de oppositie eist opnieuw dat ze een stap opzijzet “zolang dat onderzoek loopt”. 🎫 Vanavond houdt Durf opnieuw een treffen, met 1.400 inschrijvingen. Hoewel organisator Maurits Vande Reyde volhoudt dat het “geen lancering van de MR-Vlaanderen is”, komen zowel Bouchez als Jinnih Beels spreken. w16.be/p/hoe-vandenbroucke-i…

Bart Collet avatar
Bart Collet@bart·Nov 17

The assumption that a startup needs two or three cofounders was always a statement about the cost of building, not a law of nature. AI cuts that cost, so solo founders, domain experts and nontechnical people can now ship a credible v1. The part I keep coming back to: the seed-stage bar rises too. If a team of three is multiplied 100x, it can still out-build a solopreneur who is also multiplied 100x. Every 'right founder' template reflected the technology of its era. Why would this one be permanent? Which background gains most from AI: the expert, the creator or the generalist? a16z.news/p/new-founder-arch…

Bart Collet avatar
Bart Collet@bart·Nov 17

More than 90 longevity conferences already have confirmed 2027 dates, with April the peak so far. That month: AMWC opens in Monaco, London hosts The Longevity Show and the Longevity Med Summit less than a week apart, and Germany has the Forever Young Summit in Stuttgart and GSAAM in Berlin. After June the calendar is almost empty. That is most likely timing, not a quiet autumn: in 2026, October was the busiest month in the database. Planning next year's conference budget? longevents.hyperadvancer.com…

AI_4_Healthcare avatar
⟳ Bart Collet retweeted
AI_4_Healthcare@AI_4_Healthcare·Nov 16

𝑹𝒂𝒏𝒌𝒊𝒏𝒈 𝒉𝒆𝒂𝒍𝒕𝒉𝒄𝒂𝒓𝒆 𝒔𝒑𝒆𝒄𝒊𝒂𝒍𝒕𝒊𝒆𝒔 𝒃𝒚 𝒕𝒉𝒆𝒊𝒓 𝒖𝒔𝒆 𝒐𝒇 𝑨𝑰. New data from @doximity says 52% of US docs use AI daily and 85% at least weekly ... and adoption is climbing yearly. Highest daily use by specialty: - Nephrology & Oncology: 65% - Radiology: 64% - Internal Medicine: 62% - Gastroenterology: 59% Even the lowest medical discipline (Anesthesiology) sits at 31%. Common uses include correspondence, patient education, and insurance letters. Nearly half report less after-hours work -- bye-bye pyjama time -- and 81% report that AI has taught them something new. TLDR: adoption is broad, governance and verification still lag, and Shadow AI lives on. 🔗 doximity.com/reports/doctor-… #HealthcareAI

Post attachment
Bart Collet avatar
⟳ Bart Collet retweeted
Bart Collet@bart·Nov 14

9 of 154: the Middle East's share of longevity events with a confirmed date through 2027. That's the region that wrote longevity into law: Dubai set up its Longevity Authority in June. Dubai has four confirmed dates between mid-January and late March 2027. After Riyadh's Zenos Health Summit in April, the Gulf calendar goes quiet. Gulf events tend to announce one to three months out, so expect it to fill in. We updated the Gulf guide this week, every date checked against the organiser's own site. Which event are we missing? longevents.hyperadvancer.com…