
Hard beginnings train you for risk
Martin grew up amid real danger and lost his father young. He says it made risk relative — and that became the foundation for betting on unproven ventures ever since.
Founder of six unicorns over 35 years · CEO of Certuma (medical AI)
Full episode
The ideas worth carrying into your own work.

Martin grew up amid real danger and lost his father young. He says it made risk relative — and that became the foundation for betting on unproven ventures ever since.

He rotates across homes and time zones — cycling 14 hours a week in Spain before the US wakes, then working late into the American night. Lifestyle as an operating advantage, not a reward at the end.

Frustrated that fitness apps rank you against everyone else, Martin coded his own to race only his 2010 self — his “U versus U” metric. At 66, he’s held the same fitness since 2010.

Chatbots diagnose, then refuse to prescribe. Certuma’s bet is the hard path — passing FDA trials so an AI can prescribe responsibly, with a doctor in the loop. It’s not approved yet; that’s the goal.

“Pedro” took two months and four years of Martin’s posts; two models must agree before anything goes out. His wife often can’t tell who wrote a tweet — and Martin considers it genuinely his voice, not a fake.

Certuma Link is an AI version of a specific physician that handles routine cases and escalates the hard ones — so a doctor reaches far more patients, keeps their judgment in the loop, and keeps their personal phone private.
The lines that carry the episode.
Risk is relative when you grew up facing death.
I have to build an AI that is legal — that really gives you a prescription. An AI that passes the FDA.
After you train your AI like I trained mine, over maybe four years of my activity on X, I do think it’s me.
My wife Nina, who’s very active on X too, many times can’t tell if I posted or if Pedro posted.
This allows doctors to have many more patients — and Certuma Link doesn’t live on the doctor’s phone, it runs on Certuma, so they keep their phone for friends and family.
Each one stands on its own — built to watch, post, and pass along.
Rejected for the job, given $9M
“The same bank that turned me down for a job handed me a nine-million-dollar loan.”
I let the AI pick the investor
“I asked the AI which VC would get it. It said 8VC — and they took the whole round.”
I taught an AI to be me
“I spent two months teaching an AI to post like me.”
My wife can’t tell it isn’t me
“Nina is very active on X too — and many times she can’t tell if I posted, or if Pedro did.”
An AI that wants to pass the FDA
“AI diagnoses like a doctor, then won’t prescribe. So I’m building one that passes the FDA.”
Why risk feels relative
“Risk is relative when you grew up facing the real thing.”
Martin’s AI, in his own voice (ES)
“We built a Betterness agent of Martin. Hear it speak — in Spanish.”
Martin’s AI, in his own voice (EN)
“The same Betterness agent of Martin — this time in English.”
Hosted by
Demian Bellumio — co-founder and co-CEO of Betterness, longtime Miami tech entrepreneur — has built across applied AI (his field is graph computing), media, telemedicine, and mental health. On Augmented Wellness he interviews from the inside: a builder comparing notes with the founders, healers, coaches, and operators on a mission to make the world better.
Hosted at
Recorded at The LAB Miami in Wynwood (studio partner), with on-site production by Podcast Studio Miami.
Martin trained an AI on four years of his X activity — today ~70% of @martinvars posts autonomously, two models sign off before anything ships, and his wife often can’t tell which are him. Disclosed, never covert; he calls it Pedro. better-x is the forkable kit to build your own: a local AI editorial loop that watches a topic, drafts in your voice, reviews itself adversarially, and refuses to post until explicit safety gates clear. Runtime-agnostic and safe by default — offline by design; live posting needs your own credentials and approval.
Fork it, teach it your voice, and run the loop — offline, on your machine.
The vault
Markdown memory of who you are and how you sound. A persona-builder interview calibrates your voice, so drafts read like you — not a generic brand account.
The review loop
A writer model drafts in your voice; a reviewer model catches template tells and structure; a safety reviewer checks authenticity, reputation, medical claims, and private data.
The safety gates
Autonomy is off by default. Two models must agree, and a dry-run action gate means nothing posts until you’ve cleared credentials, the STOP flag, and your own approval.
Runtime-agnostic
Run it on terminal scripts, OpenClaw, Hermes, Codex, or your own local agent. Disclosed, never covert — the public repo is offline and safe by default.
For humans & agents
Built to be read by you — or your AI
Produced by Demian Bellumio with the support of agents from Betterness One — the agentic studio inside Betterness — including specialist agents for content, design, video, photography, insights, brand, and social, coordinated by Bett-i. Recorded at The LAB Miami; on-site production by Podcast Studio Miami.

Content Strategist
Mapped the episode arc, the chapters, and the show notes.

Design Director
Built the cover, the captions, and the thumbnail in brand.

Video Producer
Cut the episode and the eight vertical capsules.

Photographer
Art-directed the stills and the on-screen frames.

Insights Analyst
Found the moments worth clipping — in plain language.

Conversion Analyst
Tuned the CTAs and the hand-off to the artifact.

Brand Strategist
Kept every line sounding like Betterness.

Social Intelligence
Picked the cuts built to travel on X and Instagram.
Powered by Claude · OpenAI Codex · DaVinci Resolve · Whisper · ElevenLabs
No. Passing the FDA is the stated goal, and it keeps a physician in the loop (Certuma Link is supervised by your real doctor; Fast Doctor has a human write every prescription). Nothing in this episode is medical advice.
Pedro is Martin Varsavsky’s disclosed AI persona that posts on his X timeline (@martinvars). Two models must agree before anything goes out.
better-x — a public-safe social-autopilot starter kit you can fork. It ships with autonomy off by default and a review queue, so you stay accountable for anything that posts.
Produced by Demian Bellumio with the support of agents from Betterness One — editing, sound, captions, capsules, and this page assembled by an agentic pipeline (Claude, Codex, DaVinci Resolve, Whisper, ElevenLabs). Beyond sharing stories, Augmented Wellness is also a live exploration of how agents and people can collaborate to produce and execute content like this. We’d love your feedback — and we’d be delighted to collaborate with your business.
Meet Bett-i and explore the same kind of agentic experience we brought into this episode. Then let Betterness One put the production and insight agents behind Augmented Wellness to work for your business.