Letter to Mosaic Tiles

Dear current and future Tiles:

As we approach our first founding anniversary, Ali and I decided to share with you part of our own thought process when we started Mosaic, in hope that it helps you see what we see.

The year is 2025. We are living through a moment where machine intelligence is absorbing large portions of what we traditionally call “knowledge work”. This shift is happening so rapidly that some of the most serious people in AI are speculating that by 2028 we may have created a “country of geniuses in a data center” - an age of intelligence abundance where the essence of human labor is fundamentally redefined. If that possibility is even directionally right, we face several questions that are radically existential to human kind:

If machines can operate with human-levels of intelligence, what happens to the very nature (and value) of human labor?

How should we build new companies? How should we choose what problems to solve?

Even if they’re off by a few years, that possibility means we may be living through the last decade when what we build, how we think about it, and how we execute is within our human control. We may be looking back at this decade and realize it was our last opportunity to build consequential, authentic acts of human creation.

If this is the last window of human relevance (as we know it), then what is worth your time building?

Generational companies solve generational problems. As more and more epistemic knowledge is absorbed through machine intelligence, the world’s attention is turning towards domains where human knowledge is embodied and experiential. We believe that - out of those domains - surgery is both the most important and the most valuable problem to tackle.

To assuage any skeptics, let’s begin with a techno-economic argument. Most robots won’t outpace humans on throughput or speed for years. The tasks worth designing robots for are not high-volume, low-value tasks. They're the ones that are variable, high-stakes, and operator-dependent. The average economic output per labor hour of a household task is $30. The average economic output per labor hour of surgery is $5,000. That is a >100x difference.

But today it takes 12 years on average to train a surgeon, and 20+ years to make a good surgeon. Surgery is a craft, and crafts are inherently variable. In surgery, that variability in human skills maps onto variability in outcomes, which bears a high human and economic cost. So as everything else becomes absorbed and enhanced by machine intelligence, surgery - with its irreducible variability and its poetic opacity to machines - remains largely untouched.

This combination of attributes makes surgery one of the most valuable human skills, one that clearly benefits from the precision and reliability of robotics, but also one that is amongst the hardest to “learn”. It should be no surprise, then, that in 2026 the largest robotics company in the world - out of any industry - is a surgical robotics company. That is because robotic precision applied to surgery is an extremely valuable societal and economic output.

Today’s surgical robots, however, are built to only execute human commands. They are not equipped with the ability to learn. Without learning machines, we have no ability to digitize, encode and transfer human expertise, leaving surgical intelligence trapped in the hands and minds of individual surgeons. As a result, surgery is fundamentally constrained to individual human expertise, and may be the last $100B industry to scale sub-linearly with human input: if we want to double the number of surgeries, we need to double the number of surgeons, residents, scrub nurses, circulating nurses and surgical technicians. In a world with aging populations, rising incidence of cancer and cardiovascular disease, rising cost pressures and a shrinking workforce that is fundamentally constrained to incompressible training timelines, that equation breaks.

We asked ourselves - what would it take to make surgery more “learnable”, and therefore more reliable? This question defines the decade ahead across all domains where human knowledge is embodied and lives in perception, action, and the feedback loops that harness judgement from interactions between the mind and the environment. But we believe the opportunity to bring machine intelligence to surgery may be the single most important problem to solve in the next decade, and it holds a singular position in human civilization: it will be one of the last human skills that is structurally locked out of “intelligence”.

Ali and I believe this opportunity has been sitting in plain sight because it's hard. And because it’s hard, it deserves the world’s best minds.

The questions we outlined at the beginning of this letter are deep and heavy ones. They are questions that no small group of people should be tasked with answering.

The advantage of working in healthcare is that we get to build solutions that are undeniably human-positive. At Mosaic, we build surgical machine intelligence that is designed to work collaboratively with human operators. The goal is not to make surgery autonomous. It is to do more, with less. More access for more patients, in the hands of more precise operators, across more sites of care. With less variability, less trauma, less time in recovery, and less cost.

So if you’re considering joining Mosaic, or any other startup, I will leave you with a thought: there are really only two reasons to join a startup - one is to be able to will something important into existence; the other is to measure yourself against the world in your ability to do it. Both of those require the same thing: a problem worth solving, with people you'd trust to make the hard decisions with. Thank you, to those of you, that have chosen to build Mosaic with us.

Ad maiora semper,

Ileana and Ali