Keynote
The Future of Healthcare and AI.
Umut OnanCEO & Founder, magnus.health
The Pace of Change
A Brief Recap of
What Is Happening
umutonan.com
Perspective
How much changes in your lifetime always feels relative.
The number of innovations, and how deep and lasting they are, depends on where you stand.
Who is talking?
Ex-consultant, multidisciplinary, serial entrepreneur.
Umut Onan
Founder & CEO
M.Sc. Software Engineering & B.Sc. Economics / Computer Science — TU Wien · Munich
Consultant → Senior Manager
2006 – 2015
Built the regional digital-CX & creative practice from the ground up.
Business Design Director
2015 – 2016
Led service-design engagements across the Accenture Interactive network.
Founder & CEO
2016 – now
End-to-end product & service design for banks, insurers & telecoms.
Reach & audience
PERSONAL SITE
umutonan.com
20years of shipped work
Two decades of product and design work, first-hand.
NEWSLETTER
Innovation Insights
1,500executive subscribers
A direct line to decision-makers across banking, insurance & telecom.
Trusted to deliver for
Top insights · Biomarkers
Our blood tests and our wearables, read together.
Magnus surfaces the few patterns that matter, before they become a diagnosis.
Top insights
Biomarkers
Meals
Snap it.
Search it.
Scan it.
Search it.
Scan it.
Take a photo of the plate and Magnus recognises the meal, then fills in the calories and macros.
Every meal lands in the day’s calorie balance, next to sleep, steps and fasting.
Magnus Medical · Doctor portal
Our doctors see the same picture.
Every patient’s record and biomarkers in one place, with an AI co-pilot for the review.
Famous last words
Seeing ahead is a rare skill. Even the most successful of us often lack it.
METR · Length of tasks AI finishes alone
The work AI can finish alone doubles every 4 months.
Today2 h
+4 mo4 h
+8 mo8 h1 workday
+12 mo16 h2 workdays
+16 mo32 h4 workdays
+20 mo64 h8 workdays
+24 mo128 h3 work weeks
+28 mo256 h6 work weeks
1 square = 1 hour of a skilled professional’s work · if the trend holds
Source: METR, 2025–2026
UN Security Council · New York
23 September 2026
“It feels different now.”Sam Altman
Proteins in us
Molecular marvels that catalyze biochemical reactions.
Proteins keep our cells running, regulating molecular flow throughout our bodies.
InsulinTells our cells to take sugar out of the blood.
HemoglobinCarries oxygen from our lungs to every cell.
Serotonin receptorCatches serotonin, the messenger behind our mood, sleep and appetite.
Why it matters
Where else knowing the shape matters.
01Understanding diseaseStudy cancer mutations like p53, and the misfolded proteins behind Alzheimer’s and Parkinson’s.
02Designing new proteinsBuild brand-new proteins from scratch: to break down plastic, capture carbon, or treat disease.
03Better vaccinesKnowing the virus’s spike shape lets mRNA vaccines teach the body the right target.
One question sits under all three: what does this protein look like?
After 50 years
Most of the human proteome was still dark.
17%of the human proteome had a structure solved in the lab
5,000+human proteins had no structural information at all, not even a rough model
Source: Porta-Pardo et al., “The structural coverage of the human proteome before and after AlphaFold,” PLOS Computational Biology (2022). 17% = share of the proteome’s amino acids.
A 50-year question
The protein folding problem.
01How does the sequence of amino acids define the shape?
02Through which stages does the chain fold into its final structure?
03Can a computer predict the structure from the sequence alone?
CASP · Since 1994
The world’s protein-folding exam.
Every two years, teams predict the shapes of proteins that have just been solved in the lab, but not yet published. Then the answers are revealed.
Scored 0 to 100. Then progress stopped: for 14 years, the best stayed around 60.
Winning accuracy score
The first breakthrough · 2013
DeepMind learning without instructions.
Clip: The Thinking Game (2024)
Google DeepMind
DeepMind kept looking for bigger challenges.
The moment they found out.
The AlphaFold team
Image: Quanta Magazine
CASP · 2018–2020
Then AlphaFold broke the plateau.
2018: first entry, ahead of every team. 2020: AlphaFold 2 crosses the line of a lab experiment.
Fourteen years stuck. Broken in two competitions.
Winning accuracy score
November 2020 · CASP14
92.4
AlphaFold 2’s score. Past the line of a lab experiment.
Best before
~60
AlphaFold 2
92.4
“We have been stuck on this one problem for nearly 50 years.”
John MoultCo-founder and chair, CASP
Average error: about the width of an atom. Four years later, a Nobel Prize.
Prediction meets reality.
CASP14 · 2020
Image: Google DeepMind
Nobel Prize in Chemistry.
2024
Demis HassabisGoogle DeepMindProtein structure prediction
John JumperGoogle DeepMindProtein structure prediction
David BakerUniversity of WashingtonComputational protein design
The human proteome
Before and after.
Human proteins with no structural information
Before
5,027
After
29
Source: Porta-Pardo et al., PLOS Computational Biology (2022). “Before” includes lab structures and models of similar proteins.
Isomorphic Labs · Spun out of DeepMind
From AlphaFold to medicine.
We had been designing keys for locks we had never seen. Isomorphic Labs designs them with the lock in view.
2021Spun out of DeepMind
2024–26Lilly, Novartis, J&J partner
May 2026$2.1B raised
End 2026First human trials
Lock: the disease proteinKey: the AI-designed drug
Aug 2026Medicine · Oncology
A cancer vaccine made for each patient
42%
lower risk of melanoma coming back. The first personalised cancer vaccine to pass a Phase 3 trial.
Each patient’s tumour is sequenced, and a vaccine is built to target up to 34 mutations found only in that person’s cancer. Given with standard immunotherapy, it cut the risk of melanoma coming back.
The mRNA technology behind the COVID vaccines now makes a one-off medicine for each of us. It trains your immune system to recognise your cancer. Lung and other cancers are next.
How · Moderna and Merck (intismeran autogene + Keytruda), using tumour DNA sequencing and mRNA.
9 Jun 2026Longevity · Ophthalmology
First given to a person, in one eye.
1st
patient dosed with ER-100, a therapy designed to reset the age of our cells.
OSK: 3 of the 4 Nobel-winning Yamanaka factors, switching eye cells back to a younger state without changing their DNA. Phase 1 for glaucoma and NAION: safety first.
We have always treated ageing as a one-way street. In animals, this restored vision. Human results are still to come: the first test of whether ageing can be reversed in a person.
How · ER-100 by Life Biosciences (co-founded by David Sinclair).
May 2026Mathematics
An 80-year-old maths problem, overturned
80 yrs
A 1946 conjecture by Paul Erdős, disproved.
A long-standing assumption about how points can be arranged on a flat plane turned out to be wrong. Fields medallist Timothy Gowers said the result deserves publication in maths’ top journal.
Generations of the world’s best mathematicians had tried and failed. When a basic belief in maths turns out to be false, the work built on top of it opens up again.
How · An OpenAI reasoning model, checked by human mathematicians.
Image: Nature, 2026Aug 2026 · ScienceSynthetic biology
Working viruses, built from written code
16
working viruses, designed from scratch.
Scientists wrote complete viral genomes that don’t exist in nature, built them, and they worked. They kill E. coli, including strains that resist natural viruses.
Proteins were about reading biology. This is writing it. These viruses attack only bacteria, a new front against antibiotic-resistant infections, tied to over a million deaths a year. It also raises serious biosecurity questions.
How · Stanford and the Arc Institute’s Evo genome model.
Sep 2026Energy
Fusion plasma, kept under control
200 ms
ahead: reactor instabilities predicted, and prevented.
In tests on a fusion reactor, a control system checked the plasma every 20 milliseconds. It spotted dangerous instabilities before they formed and corrected them automatically.
Fusion is how the sun makes energy: nearly unlimited clean power. The obstacle is plasma hotter than the sun that changes shape faster than any human can react. Keeping it stable moves fusion toward power plants.
How · PACMAN, developed by Princeton and PPPL.
May 2025 → 2026Transportation
500,000 rides a week with no one at the wheel
500k
paid driverless rides a week. Aiming for 1 million by the end of 2026.
Fully driverless taxis run as an everyday paid service in more than 10 US cities, with London and Tokyo next. Driverless freight trucks began running on Texas highways in May 2025.
For decades, self-driving was “five years away.” Now we pay for rides with an empty driver’s seat, every day, at city scale. Freight is next, and moving goods is a far bigger market than moving people.
How · Waymo (taxis) and Aurora (trucks).
From Treatment to Prevention
The Silent Epidemic:
Metabolic Health
umutonan.com
Children per woman · Humanity
We are having fewer children.
We are building superintelligence, and quietly heading towards extinction.
Today
South Korea0.7
China1.0
Japan1.2
Italy1.2
United States1.6
Summer 1962
Atlantic City, New Jersey
Summer 2026
Atlantic City, New Jersey
We Are Designed to Move.
26 bones in each foot. 27 in each hand. Built to walk, run, climb and lift, for a lifetime.
Play
Climb
Lift
Today
Everything around us is designed so we don’t have to move.
Today
Everything around us is designed so we don’t have to move.
Today
If it’s in your head, it’s at your door. That’s how convenient life has become.
What’s Happening
88%
of adults are metabolically unhealthy.
And it is rising — the biggest silent epidemic of our lifetime.
454M
smartwatch users generate data they can't act on.
The data and devices exist — the understanding doesn't.
Sources: NHANES / University of North Carolina (2019); Statista, Market.us (2025).
Adults with diabetes · World
Diabetes has nearly quadrupled since 2000.
1 in 9 adults today.1 in 8 by 2050.
151M2000
194M2003
246M2006
285M2009
366M2011
382M2013
415M2015
425M2017
463M2019
537M2021
589M2024
853M2050*
Adults aged 20–79, millions. Source: IDF Diabetes Atlas, 2000–2025.*Projection
Healthspan vs. lifespan
We live longer, but not healthier.
Lifespan+6.5 years since 2000
Healthspan+5.4 years since 2000
9.6 yearslived in poor health, on average worldwide
Garmany & Terzic, JAMA Network Open 2024 · 183 WHO member states
Myopia · Prevalence
Half the world will be myopic by 2050.
4.76 bnpeople with myopia
938 Mwith high myopia
Holden et al., Ophthalmology 2016 · 145 studies, 2.1 M participants
22.9%
2000
28.3%
2010
33.9%
2020
39.9%
2030
45.2%
2040
49.8%
2050
Myopia · Collective eye test
How far you see clearly.
Scan the code — the test shows where you stand.
Reading 35–40 cm
−0.50 D
2 m
−1.00 D
1 m
−2.00 D
50 cm
−3.00 D
33 cm
−5.00 D
20 cm
−8.00 D
12.5 cm
Far point = 1 ÷ diopters
What Comes Next
Where Is AI
Heading?
umutonan.com
METR · Length of tasks AI finishes alone
The work AI can finish alone doubles every 4 months.
Today2 h
+4 mo4 h
+8 mo8 h1 workday
+12 mo16 h2 workdays
+16 mo32 h4 workdays
+20 mo64 h8 workdays
+24 mo128 h3 work weeks
+28 mo256 h6 work weeks
1 square = 1 hour of a skilled professional’s work · if the trend holds
Source: METR, 2025–2026
A different category
AI is not another invention. It changes everything, like fire and electricity did.
Leaps of our lifetime
Internet1983
Mobile phone1983
Social media2004
Cloud2006
Smartphone2007
Blockchain2009
Leaps for our species
Fire~400,000 BC
Electricity1880s
Artificial intelligenceNow
Before ChatGPT
In 2014, the people who built our technology were already worried.
“With artificial intelligence we are summoning the demon.”
Elon MuskMIT, October 2014
“The development of full artificial intelligence could spell the end of the human race.”
Stephen HawkingBBC, December 2014
“I am in the camp that is concerned about super intelligence.”
Bill GatesReddit AMA, January 2015
Where AI is heading
Three stages of machine intelligence
ANI
Artificial Narrow Intelligence
Better than us at one task: folding proteins, playing Go, reading scans.
Where we are today
AGI
Artificial General Intelligence
As capable as us at almost anything we can learn to do.
The next step
ASI
Artificial Superintelligence
Far beyond the best human minds, in every field at once.
What follows
Why ASI is different
A superintelligence would not need to hate us to hurt us.
It improves itselfEach generation builds the next one, faster than we can check it.
We cannot predict itA mind far smarter than us is, by definition, hard for us to understand.
Its goals only need to driftLike us and ants: we do not hate them, we just build the highway.
Us and the ant
An ant has no idea who we are. And we never think about it.
It cannot know what we do, what we build, or what we think. It lives in its own small world, a few metres from ours.
To a superintelligence, we would be the ant. Maybe less, as it keeps improving.
December 2015
Elon Musk co-founded OpenAI, to build AI in the open, for all of us.
Elon MuskCo-chair
Sam AltmanCo-chair
Greg BrockmanCTO
Ilya SutskeverResearch director
$1Bpledged by its backers
NonprofitNo shareholders. Research shared openly, for the benefit of humanity.
2015 → 2025
The open project became one of the most valuable private companies on Earth.
2015Nonprofit labOpen research, no owners
Feb 2018Musk leaves the boardCites conflict with Tesla’s AI
2019“Capped-profit” companyMicrosoft invests $1B
Nov 2022ChatGPT launches100M users in two months
Oct 2025For-profit corporationMicrosoft owns 27%
October 2022
$44B
Why would anyone pay that for Twitter?
Oct 2022Musk buys Twitter, renames it X
Nov 2022Purges bots; the blue check now marks paying humans
Nov 2022OpenAI launches ChatGPT
Mar 2023Anthropic launches Claude
Jul 2023Musk launches xAI
Connecting the dots backward
An AI is only as good as what it learns from. He bought the place where we write.
2022Twitter → XHundreds of millions of people, writing in real time
2023xAIBuilds Grok, an AI that reads X live
2025xAI buys XData and model in one company
2026SpaceX buys xAIAdds rockets, satellites, energy and compute
Terafab · March 2026
He offered to buy every chip they could make.
“We will buy all of their chips. I’ve said these exact words to them.”
Elon Musk, to Samsung, TSMC and Micron
Their answer: we cannot grow that fast.
Why he has to build it
Every chip factory on Earth makes 2% of what he needs.
“We either build the Terafab, or we don’t have the chips.”
Elon Musk, March 2026
What Grok, Teslaand SpaceX will need
2%Every chip made on Earth today
The size
One factory, 50 times more than every chip maker on Earth.
Every chip maker today≈ 20 GW of AI compute a year
Terafab, the goal1 TW a year · 50×
AI5 · AI6 · D3Chips for cars, robots and space
10×Tesla’s Giga Texas
1Customer: his own companies
The master plan
Most of these chips are not meant for Earth.
01 · BuildThe biggest building on EarthTerafab, Texas
02 · MakeMore AI chips than the world combinedFor Grok, Optimus and Tesla
03 · LaunchSend most of them into orbitOn Starship
04 · RunData centres in spaceAlways in sunlight, connected by Starlink
SpaceX · 2026
Roughly one launch every two and a half days.
120orbital launches
by mid-October
by mid-October
140–180planned
by December
by December
Starfactory · Starbase, Texas
Rockets, built like cars on an assembly line.
700,000 ft²Gigabay, the next building on site
≈ 380 fttall, with 24 work cells
400 tcranes lifting whole rockets
1,000 / yrStarships it is designed for
February 2026
You might remember this deal. The goal: grow AI in space.
Power from the sunAlmost constant sunlight, no grid needed
Cold of spaceChip heat radiated away, no cooling plants
Closing
Closing Words
umutonan.com
Anthropic · Revenue concentration
46%
of all spend comes from the top 1% of customers.
Last August25%
Now46%
Thank you.
Your feedback and ratings are truly appreciated.
Please scan the QR code to leave your vote and comments — it only takes a minute.
Umut Onan
CEO & Founder, magnus.health
References
Videos
01
Video reference
02
SpaceX Moon and Mars bases
03
Video reference
YouTube Shortsyoutube.com/shorts/kOCEbBxH44w
04
SpaceX Makes History Catching Rocket on Landing
Bloomberg Tech · Oct 2024youtube.com/watch?v=RYUr-5PYA7s
05
Video reference
06
Video reference
TikTok · @spacexlunchstiktok.com/@spacexlunchs/video/7694371687499828493
07
Starlink train seen from Earth
08
Super Heavy engine ignition update
09
Moon, Optimus and data centre renders
The Space Racethespacerace.news
10
Video reference
YouTube Shortsyoutube.com/shorts/VrO9uAHN0po
11
Video reference