Healthcare AI is getting better at identifying risk, and bringing the right partners together can turn that capability into a service. AI can find the signal, but healthcare needs infrastructure to act on it. That is what Life AI is building with Life Cloud, a connected healthcare network, around one principle: bring intelligence to where care happens. Care happens in hospitals. They are where clinical knowledge originates, and they hold what turns a signal into care: laboratories, specialists and clinical judgment. But a patient’s path rarely stays within one hospital, and prevention can stall between signal and care, often in the handoffs between hospitals. Closing that gap takes two things: intelligence that works inside each hospital, and a network that connects them. In Life Cloud, each hospital joins as a node. The AI flags; the doctor makes the final clinical decision. Thalassemia screening, one of Life Cloud’s applications, shows what that involves. It is one of the most common inherited disorders worldwide, especially in Southeast Asia, and severe forms require lifelong treatment. Carriers usually have no symptoms, so the risk often goes unnoticed until a child is diagnosed, though it can be identified years earlier. A routine blood count, where AI can flag signs of the trait, is only the first step. Confirmatory testing, genetic counseling and partner screening must follow, often at different facilities. Partner screening is often the most fragile link. For health systems, things may come to those who wait, but only the things left by those who hustle. Where that sequence is connected, families can be reached in time. In Guangxi, China, an information system tracks each high-risk couple, and screening coverage reached 99.99% in 2024. One remaining gap, a 2026 review notes, is linking systems across provinces. Life Cloud is designed to connect the institutions patients move between. Models must also hold up across populations. Carrier rates in Vietnam range from under 10% among the Kinh majority to 40% in some ethnic groups, so a model validated in one group must be checked in others. Every hospital that joins makes that possible for more people. The aim is not to decide for families, but to give them accurate information and counseling early enough to make their own choices. Life Cloud is at an early stage, in closed testing with a few hospitals, and the thalassemia application is in development. It will be judged in care, by whether the pathway can be completed reliably. Longer term, a network this diverse could support research into how medicines work in Asian populations. Life Cloud will grow one node at a time. Each hospital, research institution, pharmaceutical company and technology partner that joins extends what the network can do for patients. We look forward to building it with those addressing similar challenges.

What We Know About Disease Matters. What We Do With It Matters More. In Vietnam, more than 14 million people carry genes associated with thalassemia. Many appear healthy and may not know they are carriers. A routine blood count can offer the first clue. Genetic counseling can help families understand what that means, and screening has sharply reduced severe cases elsewhere. Yet around 8,000 children are still born with thalassemia in Vietnam each year, including 2,000 with severe forms. Finding a risk is only the beginning. What follows often matters more. Stroke is different, but it raises a similar challenge, with higher stakes. Risk builds over years from genetics, blood pressure, diabetes and more, and no single test tells the whole story. In a study of 2,931 Chinese adults, adding genetic information to a clinical model improved the classification of 20-year stroke risk by 17.2%. That does not mean 17.2% fewer strokes. It means the model distinguished risk more accurately. Validated for the people it serves and read with clinical judgment, that insight can help prevention start earlier. This thinking helped shape StrokeGENME, developed with Kalbe Farma and KalGen Innolab in Indonesia. KalGen brought laboratory expertise. Life AI contributed genetic-risk interpretation technology. Kalbe brought its healthcare network. Together, we developed a service intended to help people understand their inherited stroke risk earlier and give clinicians another piece of evidence to consider when discussing prevention. There was another lesson in how StrokeGENME came together. Because each partner brought what it had already developed, the program took three months rather than an estimated 24, at an estimated 10–30 times lower development cost, freeing resources that can go back into patient care. What got faster was not biology. It was the assembly. That distinction matters beyond stroke. Across Asia, hospitals, laboratories, researchers and pharmaceutical companies already hold valuable knowledge and expertise, and new AI tools and genomic discoveries continue to add to it. The opportunity is considerable. So are the demands: pathways are long and regulated, and turning knowledge into care requires scientific validation, clinical understanding and work across organizations. StrokeGENME showed us what becomes possible when the right capabilities come together. The next challenge is making that process easier to repeat, across more partners, more programs and more healthcare needs. At Life AI, we see our role on both sides of that work: providing technology partners can build on, and infrastructure that lets each new program start from what is already validated. Every new application will bring its own challenges, but met responsibly, each can help more of what medicine knows reach the people who need it. AI will help us know more about disease, faster. Are we organized to act on what we know?

Can a small fish outside Silicon Valley swim with the giants? On October 20, OpenAI bought Orchestra Bio (Congratulations to @ashokaraj_), whose agents run the unglamorous side of drug development: budgets, timelines, scheduling. It joins Rosalind, OpenAI’s life-sciences suite. Alone, a small deal. With the rest of the year, a pattern. Since January: NVIDIA and Lilly, up to $1 billion for a shared AI lab. Anthropic, a reported $400 million for six-person Coefficient Bio, then its own wet lab. Isomorphic Labs, Demis Hassabis’s drug company, $2.1 billion. Google and Meta into Biohub’s $1.8 billion push. Anthropic went to the bench. OpenAI went to the calendar. DeepMind went after the molecule. Life AI started on a shared bench at UCSF and now works with hospitals across Southeast Asia. On paper, we are the fish that gets eaten. My take: this is the best news we have had all year. First, the tools became rentable. Rosalind and Claude are services, Boltz-2 is a free download, and equivalent AI capability gets about ten times cheaper a year. The giants’ billions are becoming our cheapest line item. Second, capital noticed. Every deal above is a price tag on AI for biology; investors now want anyone with real patients and real data. Third, most important: the giants compete with each other, not with us. To win, each needs what none has: patients, hospitals, biology from the places their training data forgot. The faster they race, the more they need partners. Now the fair part. Orchestra and Coefficient were acquired because they built horizontal software: tools any lab could use and any giant could build or buy. If your only asset is a model or a workflow, you are a feature waiting for an acquirer. A mission does not survive an acquisition the way a cap table does. Life AI exists for the populations medicine keeps leaving behind, Asians most of all. No frontier lab will buy that mission and keep it. A small fish cannot beat a big one in open water. It survives on reefs, in shallows, in schools. Biology has the same geography. Which brings me to DeepSeek. In January 2025, a lab in Hangzhou released a model that matched the frontier, built on a base model whose final training run, it said, cost about $5.6 million. NVIDIA lost $589 billion in a day. This is about method, not politics. DeepSeek treated constraints as a design brief, opened what it built so the world adopted it overnight, and refused to play the giants’ game. A DeepSeek of life sciences will not be a cheaper AlphaFold. Compute is not biology’s bottleneck; data and the clinic are. It will be a lab outside the valley that trains on people the giants’ models have never seen, validates in hospitals they cannot reach, at a tenth of the cost, in the open. That lab does not have to be Life AI. But it will come from where we are, not where they are. Small fish do not survive by out-swimming sharks. They survive by knowing the reef. I admit I’m biased. I’m a small fish.

Grateful to have been part of the Google for Startups Accelerator: Southeast Asia 2026. We brought Life Cloud, our AI-powered cloud for healthcare, to Google San Francisco for a week of building and stress-testing with Google engineers. We're taking those lessons back to our work with health partners across Southeast Asia. Thank you to Google for Startups and our mentors, and congratulations to the 23 other startups in the cohort. #GoogleForStartups #HealthcareAI

Thank you, @GoogleStartups, for a remarkable week in San Francisco. We brought Life Cloud, our AI-powered cloud for healthcare and one piece of the infrastructure we're building for drug development. After a week of stress-testing with Google engineers, we're heading home with sharper answers for our health partners across Southeast Asia. Proud to be part of this cohort.

Clinical Development Needs Shared Trust, Not Just Another Central Coordinator Clinical development has a paradox: it takes many independent organizations to run a program, yet coordination and records are often managed through sponsor-led or program-specific systems. The sponsor, often working with a contract research organization (CRO), oversees the program and coordinates its activities. When a subsequent program involves different organizations, prior records, partner qualifications, and approval workflows are not automatically reusable. Teams can end up rebuilding parts of the same operational groundwork. A better coordinator alone does not solve this problem. When records depend on one party’s control, other participants must decide whether and how to rely on them. A decentralized architecture can reduce that dependence by allowing participants to share responsibility for maintaining and verifying records under jointly agreed rules. Decentralization changes who maintains the record and sets the rules, not who does the work: The record of who performed and approved each step is maintained and verified jointly. The rules for how programs run are agreed by the participants and applied consistently to all of them. The work stays with each organization, which carries out its own tasks without surrendering operational sovereignty. This does not replace governance. It must operate alongside clear institutional responsibilities and shared definitions of progress. When it does, the record and rules no longer belong to any single program, allowing clinical programs to move from bespoke projects to reusable rails: Network value: Every program adds partners, validation capacity, and verified contributions to the network, making subsequent programs faster, less costly, and easier to coordinate. Accountability: Every result remains traceable to those who produced and approved it, and verifiable by every actor that relies on it. In clinical development, the challenge is making coordination reusable across programs with distinct protocols, institutional responsibilities, and site-level integration. Life AI helps design, validate, and deploy clinical programs. Its shared infrastructure and coordination layer is designed to reduce the need for repeated bilateral negotiations, bespoke compliance work, and site-by-site integration. The goal is to make operational groundwork more reusable across programs through a neutral protocol, while each organization retains its own regulatory responsibilities. What would it take for competing sponsors to rely on the same verifiable record?

Is Life AI Biohub a subsidiary of Chan Zuckerberg Biohub? A couple of VCs messaged me that question last week, right after the news that the US government, Google and Meta are joining Biohub in a $1.8 billion push for AI-powered biology. The answer is no. (Even though I wish it were.) I understand why they asked. Open lifeai.io and you will find a lot of Biohub in it, from the vision down to the way we execute. I will confess it: we were deeply inspired by them. Life AI started eight years ago as a research project inside UCSF. Our lab was a shared bench, where we ran our very first study on the genomics of Asian populations. About ten minutes away, Chan Zuckerberg Biohub was pursuing an extraordinary ambition: helping cure, prevent or manage all disease by the end of this century. The mission deeply inspired us. We borrowed most of that mission. Then we narrowed it: chronic diseases like diabetes and cardiovascular disease, for the populations medicine keeps leaving behind, especially Asians. In those early days we got a lot of support from Prof. Joe DeRisi, Biohubs founding co-president from UCSF. I doubt Joe remembers us: three people, no funding, don’t even have their own bench, but with a big dream of walking the same road as a $600M Chan Zuckerberg Biohub. But that is where Life AI was brewed, inside UCSF and immersed in Biohubs culture. Now, back to the $1.8 billion. Most coverage of the Virtual Biology Initiative stops at the virtual cell: an AI model that predicts how a cell responds to a drug, a genetic edit, or a change in its environment. Biohubs head of science @alexrives expects a first dataset in about a year and accurate predictive models in roughly five. The headline everywhere: faster drug discovery. I dont think the Virtual Biology Initiative stops at the cell, or just for drug discovery. The next model is the virtual human, and the place it matters most is the clinical trial. Here is the uncomfortable math. AI is making discovery faster and cheaper, so more candidates will reach the clinic. Trials run on people. About 90% of candidates that enter trials fail. So most of the time, real people carry real risk, including serious side effects, for a drug that never makes it. AI has raced ahead, yet humans are still the test objects. I believe the real legacy of the Virtual Biology Initiative will be ending that. I admit Im biased. Im human. Personal note: Mark Zuckerberg @finkd and Priscilla Chans podcast episode How AI Will Cure All Disease is fantastic. Highly recommended. Links in the first comment.

The next phase in the evolution of networks will be shaped by what they can coordinate, not just what they can verify. Stablecoins improved how value moves between parties. Real-world assets made ownership more transparent and programmable. Healthcare AI presents a different coordination challenge: aligning technology companies, healthcare providers, experts, communities, patients and autonomous agents around shared outcomes. Blockchain can provide auditable records of onchain ownership and transactions, while AI agents are beginning to execute tasks across systems. Around this infrastructure, Web3 is developing trust primitives such as onchain identity, attestations and agent reputation. These primitives can establish provenance and record participation, but they do not by themselves determine how contribution should be attributed or evaluated within a specific context. This creates a new coordination question: how can networks attribute and evaluate meaningful contribution across participants? This shift from Proof of Ownership to Proof of Contribution is the principle behind Proof of Serving Life (PoSL). PoSL is currently applied through Life Unity, Life AI’s community layer, where contribution is evaluated across three dimensions: Proof of Work reflects verified activity and community-building efforts. Proof of Commitment considers the depth and duration of participation. Proof of Impact evaluates whether those efforts translate into real usage and observable outcomes within the ecosystem. Together, these signals connect activity, continuity and outcomes. The objective is not to create a universal reputation score, but to make contribution more structured, attributable and explainable within a defined context. As healthcare AI becomes more collaborative, networks will need more than proof of identity and records of activity. They will also need a reliable way to understand how different participants contribute to shared processes and outcomes. Blockchain made ownership and transactions verifiable. AI is making action autonomous. As healthcare AI evolves, making contribution legible becomes an important coordination challenge.

Healthcare AI will not scale if every program has to rebuild the same coordination layer. The opportunity is to move from bespoke projects to reusable rails. Across healthcare and drug development, every program must coordinate a complex operating layer: institutions, clinical evidence, validation workflows, regulatory pathways, privacy requirements, access rights and deployment partners. When that operating layer is rebuilt program by program, execution remains slow, fragmented and difficult to scale. For Life AI, shared infrastructure becomes the reusable rail for coordinating AI, biotech, clinical validation and real-world deployment across programs. But “reusable” does not mean every program uses the same model or workflow. A stroke prevention program, an oncology workflow, a metabolic disease program or an autism intervention study may each require different domain logic, partners, evidence standards and deployment paths. What can be reused is the infrastructure underneath. These rails appear across several core infrastructure components: Purpose-built world models for domain-specific healthcare programs. Multi-organization infrastructure for privacy, ownership and access transparency across multi-party healthcare environments. Clinical validation pathways for moving from insight and study design toward evidence, validation and real-world deployment. Coordination primitives for identity, role, capability, permission, evidence, decision-making, accountability and feedback loops. Together, these components define how participants enter a program, what they are allowed to do, what evidence supports their work, how decisions are made and how outcomes return to the system.a Over time, each deployment can strengthen the next. More partners, evidence, validation workflows and real-world feedback can accumulate into shared infrastructure. That is how healthcare AI moves from isolated deployments to infrastructure that compounds across programs.

Nobel season usually makes me optimistic about what science can achieve. Last year’s prize honored discoveries from 1995 and 2001. GLP-1, the hormone behind today’s obesity drugs, was identified in the 1980s, and the first drug in its class was approved in 2005. Even AI, the technology everyone calls the future, won a Physics Nobel in 2024 for neural network research that began in the 1980s. Most of what we call breakthroughs are old discoveries that finally got proven. Drug development runs on the same clock. It takes 12 to 15 years to get from idea to approved medicine, and about nine in ten candidates that enter human trials never make it. AI is now speeding up drug discovery, and that is real progress. I’m an AI founder, so you might expect me to say AI solves the drug development problem. I don’t think it does. Discovery is not what medicine is short of. In one analysis, AI-discovered molecules passed Phase I trials 80 to 90% of the time, well above the industry norm. In Phase II, where a drug must show it works in patients, 4 of 10 succeeded, about the industry average. I’ve seen this up close. At Google I learned how far good infrastructure carries something once it works. In life science and healthcare I learned the harder lesson: a result can be correct and still go nowhere, stuck between the sample, the lab, the report, the clinician and the next action. So when discovery accelerates, how does scientific progress become human progress? If AI hands us 100 times more discoveries and the same capacity to prove them, we don’t get faster medicine. We get a longer waiting list. This Nobel season, I am thinking less about how many discoveries AI might produce, and more about how many years of waiting it might remove. On a personal note, I have been following @JeffDean's work, first at Google and now at Discovery Loop, and @DaphneKoller’s at insitro. I hope what they are building takes years off that wait. Who knows, they may win a Nobel Prize in 2030 for the work they are doing today. That would be the best proof that the wait is getting shorter.

.@CoinMarketCap we would like to register business@genefriendway.com to manage Life AI's self-reporting dashboard on CoinMarketCap.

Infrastructure only matters when it changes what organizations can do together. One example is our work with Kalbe Farma and its subsidiary, KALGen Innolab, on StrokeGENME, a genetics-powered platform that translates DNA insights into individualized stroke-prevention plans, delivered through Kalbe’s nationwide ecosystem in Indonesia. Bringing StrokeGENME from genetic insights to real-world healthcare delivery brought together capabilities that already existed across the partnership: KALGen’s laboratory capabilities, Kalbe’s healthcare expertise and reach, and Life AI’s technology platform. Kalbe did not need to become a technology company. Life AI did not need to build a clinical network in Indonesia. KALGen already had the laboratory capabilities required on the ground. No single organization needed to own the entire pathway. Each organization could contribute what it was already good at, bringing an existing capability into the same pathway. That leverage changed the path to deployment. The program has since reached real-world deployment, with timeline reduced from approximately 24 months to 3 months, and was subsequently accepted by regulator. That combination is central to how Dr. Tuan Cao, Co-Founder and CEO of Life AI, described the partnership: “Our partnership with Kalbe amplifies impact, pairing AI with a healthcare leader built to reach millions.” The challenge is making those capabilities work together without rebuilding the path from scratch every time. How much of the coordination between technology, laboratories, clinical expertise and healthcare delivery can become reusable? Because if every new collaboration starts from zero, it remains a project. When the path itself becomes reusable, partnership starts to become infrastructure. That is one of the questions we are continuing to build around at Life AI.

.@CoinMarketCap, we would like to register business@lifeai.io to manage Life AI's self-reporting dashboard on CoinMarketCap.

@AnthropicAI , a company whose entire product is intelligence, just confirmed it runs a wet lab. Within a day, several of our investors had pinged me with the same two questions. Is this a trend? Will every frontier AI lab end up running its own wet lab? And if so, is Life AI losing its competitive advantage? Short answers: yes, and no. Here are the longer ones. I understand why Anthropic did it. I have a fair idea how that internal debate went, because we had the same one eight years ago. Back then it looked like an odd move for a company that started on the software side of genomics: we built a physical sequencing center, the largest in Southeast Asia, designed for 500,000 genomes a year, with all the freezers and sample logistics that come with it. We did it for the reason Anthropic gives now: in biology, the final test still happens at the bench. A model can propose. Biology doesn't have to agree. So to the first question: yes, expect more of this, and the reason is arithmetic. The marginal cost of a hypothesis is heading toward zero. The cost of a failed trial is not. Roughly nine in ten candidates that reach the clinic still fail, and ten times more candidates does not fix that ratio; better evidence before the trial does. A company that sells intelligence has just concluded that intelligence is not the bottleneck. That is the real news, and every model company will reach the same conclusion once its predictions meet a petri dish. Anthropic seems to know this. It says the lab is focused on fundamental biology, not drug discovery, and it doesn't want to be seen competing with pharma. That is structure, not modesty: pharma pays its bills, and a model company that starts producing evidence for approval competes with its own customers. So the lab exists to make a model better at biology. Its product is a smarter Claude. Our lab was built for a different job. Its product is evidence a regulator will accept. A research-grade experiment tells you whether a hypothesis is interesting. A regulatory-grade one tells you whether a patient can be treated on the strength of it. Evidence at that grade cannot be rented by the token. It takes a robotic wetlab, validated processes, and a loop that runs from study design through validation to deployment, and we started building ours before models were producing candidates at today's volume. Which brings me to the second question. No, I don't lose sleep over Anthropic, DeepMind, or whoever builds the next wet lab. A better model is the best supplier we could ask for. Every jump in capability sends more candidates toward the narrow door where proof happens. We spent eight years building the infrastructure for this moment. Welcome to the bench, Anthropic. There's plenty of room. To the investors who asked, and to any who want to join the conversation: the question I'd put back is not which model will win drug discovery, but the next step. Who can prove what the models find will be the final winner.

Today in Singapore, we took a question we have been working on at Life AI to a much bigger room. As AI accelerates drug discovery, how do we build the operating infrastructure to move what we discover through clinical validation, evidence, and real-world execution? Our Co-Founder & CEO Dr. Tuan Cao shared Life AI’s perspective on that question today at the NVIDIA Inception Grand Challenge Finale. And much of the value lies in what follows: new perspectives, sharper questions, and meaningful connections across the industry. Singapore extended the rails we’ve been laying, bringing our work into a broader conversation across AI, drug development, and healthcare. And that is exactly where we want to be.

HUGE

The countdown to the NVIDIA Inception Grand Challenge Finale is on. Tomorrow, our Co-Founder & CEO Dr. Tuan Cao will take the stage as one of the Top 20 finalists, sharing Life AI’s approach to building the operating infrastructure for AI-enabled drug development from accelerating discovery to clinical validation and real-world execution. 📅 September 22, 2026 | Singapore If you’re joining us in Singapore, we’d love to connect.

What kind of operational infrastructure does coordination across the healthcare value chain actually require? Eight years of building clinical programs across Southeast Asia, working with governments, hospital networks, and pharmaceutical companies, produced one consistent finding: the challenge is not the absence of capable AI. It is the absence of operational infrastructure that makes coordination reusable across programs. Without that infrastructure, coordination remains highly program-specific, especially in drug development, where multiple key players must work together across the development process. The infrastructure this requires has three properties. It has to be reusable. Clinical programs should move from bespoke projects to reusable rails, with shared coordination rails replacing bespoke integration at every site. It has to compound. Every application should strengthen the model, the network, and the protocol, building greater capacity across the ecosystem over time. It has to enable coordination among independent actors. Hospitals, doctors, labs, pharma sponsors, regulators, patients, and AI builders can contribute services, validation, and clinical execution without surrendering operational sovereignty. At Life AI, we are building an operating infrastructure for drug development around these requirements, bringing together AI-driven discovery, wet-lab screening, and clinical validation.

Why is validation so difficult to accelerate? Because unlike discovery, validation cannot be completed in isolation. An AI-generated drug candidate still has to move through key players across the healthcare value chain, including pharma teams, clinical sites, hospitals, clinicians, patients, and regulators. Each operates with different timelines, evidence requirements, compliance constraints, and operational priorities. A candidate can move forward on one front while remaining constrained elsewhere, whether by patient recruitment, study execution, evidence generation, or the decisions required to advance it. This makes validation more operationally complex than discovery. Discovery can be accelerated within a model, platform, or controlled environment, while validation depends on how effectively these key players coordinate and execute across the development process. With more candidates entering development, that operational complexity compounds, requiring more evidence generation, clinical execution, patient participation, and coordinated decision making across organizations. As AI accelerates discovery, the need for better coordination and the infrastructure to support it becomes increasingly important downstream.

AI is accelerating drug discovery. But the pressure is shifting downstream, toward the work required to validate what discovery produces. Across science and life sciences, AI is moving deeper into discovery. Anthropic is expanding AI into scientific research, Isomorphic Labs is scaling AI-first drug design and development, and Discovery Loop is building systems to automate experimental loops. As these capabilities advance, more targets can be explored, more molecules can be designed, and more potential candidates can be generated. The discovery layer is becoming faster and more expansive. But accelerating discovery does not automatically accelerate the path to validation. A promising candidate still has to be evaluated, tested, and supported by sufficient evidence before it can move forward. Clinical validation is where this downstream pressure becomes especially visible. The gains from faster discovery can begin to narrow if the path to validation remains slow and difficult to scale. For Life AI, this raises a critical question: How do we make sure the path to validation can keep pace as AI accelerates discovery?