I run an AI company, and we just closed our Series B. I want to describe what building feels like right now, because I have not seen a moment like it. Most days it is the best job I have had. Some days it feels like pouring a foundation on ground that keeps shifting. Both are true at once.
Start with what is now possible. A small team can build what used to require hundreds of people and several years. Tools that did not exist three years ago let a handful of engineers ship products that once demanded a full company. For most of business history, the binding constraint on a startup was resources, mostly money and people. That constraint has largely fallen away. You can build almost anything now. Deciding what is worth building is the hard part.
Then there is the pace. The technology we are built on improves every few months, sometimes faster. A capability that was a research demo last year ships as a standard feature this year. The floor keeps rising under everything we make, which is thrilling and unsettling at once. A plan that looked sound in spring can look dated by fall. You learn to hold the roadmap loosely and the conviction tightly.
There is a popular theory that in this environment taste becomes the edge, since anyone can build. I think the sharpest version of that argument actually undercuts it, and it is worth borrowing: judgment commoditizes the same way everything else AI touches does. When every team can build well, building well stops setting you apart. Good judgment is not the moat. It is the price of admission. The harder question is what separates you once everyone has cleared that bar.
In my industry the answer is trust, and trust cannot be rushed.
Let me be precise about what is actually getting solved, because the hype obscures it. The near-term wins are unglamorous. Summarizing records. Reducing documentation burden. Helping a clinician navigate complexity without reading two hundred pages first. Intelligence, it turns out, is the easy part. We have plenty of it now, and more arrives every quarter. The hard part is trust, and in healthcare trust is the only currency that matters.
Here is the trap that catches people. Connectivity is not the same as usefulness. In my industry records move freely now, more freely than ever. But they still do not reliably turn into something a clinician can act on, and that gap is where the work lives. Moving data is solved. Making it trustworthy is not. A clinician will not act on a number she cannot trace to its source. Source traceability is not a compliance checkbox. It is the thing that earns the right to be used at all.
This is where the speed I described collides with a regulated industry, and the collision is the most interesting part of my job. Fast is not the same as compliant. We hold HITRUST certification and a SOC 2 attestation, the frameworks that prove our systems handle sensitive health data securely and that every control can be audited. They exist to slow you down at the exact moments you most want to move, and they are right to. A capability I could ship in a weekend for a consumer app takes months in healthcare, because shipping it means proving it, documenting it, and standing behind it when a patient’s care depends on the output.
So I live inside a contradiction. The tools beg me to go faster, and the domain demands I go carefully, and both pressures are legitimate. The mistake is to treat compliance as a tax on speed, a brake a better team would not need. It is the opposite. The constraints are the moat. A competitor with quick hands can copy a feature in a quarter. It cannot copy years of earned trust, validated accuracy, and a record regulators and clinicians have learned to rely on. When everyone has the same models, the slow, unglamorous, hard-to-rush work is the part that does not commoditize.
All of this changes the job of running the company. When resources stop being the bottleneck and the tools keep reinventing themselves, the scarce thing is judgment about where to spend the speed you have and where to refuse it. The strong founders can commit to a direction while everyone else chases the newest release. They can also tell the difference between a problem to solve in a weekend and one they have no business solving fast. Most of the thousand newly possible things are distractions wearing the costume of opportunity.
None of this is comfortable. You bet the company on a technology that can shift in a quarter, inside a regulatory frame that moves in years. The talent you want is being fought over by firms with deeper pockets. A great deal of what gets called AI is theater. Leading through that takes a steady stomach. I would not trade it for an easier moment, and I do not know a founder who would.
The rules of building a company rarely change all at once. They are changing now, faster than before. The founders who will matter in ten years are building in the middle of it, while it is still loud and unsettled. In my corner of it, the lesson is that the parts you cannot rush are the parts worth owning. Waiting for the dust to settle is one mistake. Treating speed as the only virtue is the other. That is the job right now, and there has never been a better one.
