Bart Collet

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

[Bart Collet @bart]

18 platforms checked against their own terms. None of the AI app-builders sign a BAA.

Lovable, https://t.co/W783L9rpvx, 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?

https://t.co/l2jXF8Ha7d

[Bart Collet @bart]

25,000 people are on the waitlist for a $499 scan at Neko Health's New York clinic. England's NHS waiting list is 7.33 million treatments, 111,000 of them over a year. One queue reads as demand, the other as failure; the difference is not the length but whether you chose to stand in it.

The same inversion runs through AI. Blue Cross Blue Shield links hospitals' AI coding to $942 million in extra costs from 2023 to 2025, about 70% of it from diagnoses that changed no care. Meanwhile retailers pull out self-checkouts and clinics install AI receptionists.

Does your next AI purchase give a patient more of a person, or a process?

https://t.co/avS2lmgMSV

[Keith @keithdunn]

In 2022 they stopped together after 101 yards.
In 2024 they stopped together (along with Frank Gielen) after 110 yards.
These guys are good friends. Of course they were going to stop together this time as well, after setting a new world best.

[numeriKare.be @numerikare]

Health Tech Sessions in Leuven: "Europa heeft innovatie, maar moet sneller schakelen"  - https://t.co/SeME3wSpBq @Leuven_MindGate @BEMEDTECHTWEET @bart

[Bart Collet @bart]

#backyardultra 5 days and 3 hours

[Bart Collet @bart]

Still running 😱😱 #backyardultra

[Пьяный сосед в лесополосе @donkiton24]

@bart Thank you so much, Bart! Really appreciate you including RemCard in Clinicians That Code and supporting clinicians who build their own tools. It’s great to see my project recognized by the community! I’ll definitely add the badge to the GitHub README and website.

[Bart Collet @bart]

$45,000 a year is the top of the hospital longevity market. Northwestern, Weill Cornell and Mayo Clinic all sell cash-pay longevity care: Weill Cornell charges $12,000 a year, $7,500 for a baseline, MRI on top.

Weill Cornell's FAQ gives the motive away. The goal, it says, is to prove the clinical value of longevity care so insurers eventually reimburse it. That is a trial with the money running backwards: in a study the hospital pays the patient; here the patient pays the hospital.

A GLP-1 drug with the SELECT trial behind it cannot show an insurer savings. What is a whole-body MRI proving?

https://t.co/G6aBMUEHEl

[Bart Collet @bart]

Bronze 🥉#backyardultra Belgium 🇧🇪

[Bart Collet @bart]

#backyardultra Belgium still running. 114 laps 😱😱😱

[Keith @keithdunn]

After 96 yards at Big’s Backyard Ultra World Team Championship, four nations remain in the race, each with two runners. The USA is in first place with 1099 yards, followed by Japan (1021), Belgium (1002), and France (880).

[Bart Collet @bart]

We track 500+ longevity events, and for most of its history the market behind them focused overwhelmingly on men. Women's biology needs a fundamentally different approach: different risks, different timing, not a smaller dose of the same answer.

The money started noticing in Q1 2026. New Economies mapped 115 consumer longevity startups. Jane Fonda and Mark Hyman were named to headline a women's health summit in San Francisco. Christina Farr called it 'the massive opportunity in women's longevity'.

The science was not the bottleneck. Attention was.

Who is building a serious women's longevity company?

https://t.co/Rr1lusJZky

[Bart Collet @bart]

Hospital AI rarely reaches the bedside: the plumbing (security, access, connections) takes months.

Researchers turned it into reusable code. Deploy: 5 minutes.

Their tool caught a serious inconsistency in 1 of 13 documents, 98% confirmed by doctors.

https://t.co/yp6F6ZPqxQ

[Bart Collet @bart]

@HoomanGoesWild I see a lot of companies building adjacent products to the current vibe-coding ones: scaffolding, security, compliance, etc

[Bart Collet @bart]

One engineer built Stripe's Harbor v1. Two engineers built Ramp's Glass in a few weeks.

That reshapes build versus buy for internal vibe-coding tools. Lovable and Replit exist, but pointing them at internal data is risky, and matching a design system is hard. A small in-house tool does both.

The side benefit: engineers get real practice running AI systems while non-engineers ship their own tools.

The open question: who maintains all those apps in two years? Does a one-engineer investment stay that small?

https://t.co/cGb8iIH1RJ

[Bart Collet @bart]

In 2000 Enron and Blockbuster announced a 20-year deal to stream films. Before one movie reached one customer, Enron had booked more than $110 million of profit from it. The venture existed to be announced; eight months later it was dead.

That is one of six unusual partnerships here, and none of them died of being weird. Nintendo and Sony killed their console on stage, over who owned the revenue. WWF and NBC launched a football league on a deadline instead of a customer.

The unusual pairing was never the risk.

Which of the five failure modes is in your partnership pipeline right now?

https://t.co/LAhlX9MWSH

[Zane Durante @DuranteZane]

1/ Can AI help doctors make more informed decisions based on patient activity and behavior over the course of a hospital stay?

Today we introduce TRACER: a training-free multimodal harness that enables LLMs and VLMs to track patient activity using bedside video and health records.

Blogpost: https://t.co/f7Wy0tn3df

Paper: https://t.co/5qeoDjwM7y

[Sid Sijbrandij @sytses]

We’ve just made public 125,091 DICOM radiology images from across my cancer journey (link below). The data is both longitudinal and multi-modal, dated and linked to my treatment timeline, so you can compare imaging across initial diagnosis, standard of care, first remission, “founder mode cancer care”, and second remission. It’s viewable in OHIF, the standard open-source viewer used in radiology research.

Given the long history of applying deep learning to radiology, we’re hoping to see a variety of methods and models applied to my data. If you find something interesting, please reach out to [email protected], or if you know a researcher who should see this, tag them below.

Thanks to Terra LaFerriere for organizing these images, and Kamil Slowikowski for building https://t.co/ek0DxOgvEV into the resource it is.

[Bart Collet @bart]

Tesla did not start with an affordable car. It built the Roadster for people who could pay, and that revenue funded the Model S and Model 3. The cheap product came second, on purpose.

Preventive healthcare keeps trying to skip that order. A premium preventive programme only works as a first stage if the early margin is reinvested into making the service cheaper. Most keep the margin as the business.

Tesla's other move is harder to copy: in 2014 it open-sourced its patents, betting that a bigger category helps the leader more than hoarding does.

Start premium, share the IP, sell direct: which of these is your programme actually doing?

https://t.co/KOVtXC3aD9

[Bart Collet @bart]

118 breakout attempts, nine frontier models, zero escapes. That is Perplexity's claim for SPACE, its newly open-sourced agent sandbox.

Models reportedly did escape other sandboxes that were not open source. Open code is not automatically safer, but it lets outsiders try to prove the claim wrong.

Caveat: 118 attempts is a small sample, and the test is Perplexity's own. Zero escapes so far is not the same as unescapable.

Who should run the adversarial testing: vendors, customers or an independent body?

[Bart Collet @bart]

Tesla sold a premium Roadster to fund the Model 3. Preventive healthcare could do the same: launch advanced longevity services for early adopters, then use the revenue to bring them to everyone.

Full read: https://t.co/kTFU1C9O5K

#HealthTech #Longevity

[Claude @claudeai]

A cardiology fellow built a 3D tool to help trainees learn to read coronary angiograms.