01
AI fitness partner
An AI training partner that programs workouts, tracks progress, and gives a trainer leverage over more clients.
health and fitness
gym members, gyms, personal trainers
5 products live · shipping today
We are an AI native company that builds AI native products.
Instead of generic chat interfaces or jack-of-all-trades tools, we deliver deep, domain-specific workflows engineered for high-stakes professional work.
We build AI agent powered applications to equip builders and domain experts with AI-native workflows purpose-built for their field. Moving beyond simple conversational UI, our tools deliver deep, specialized automation where accuracy, context, and mastery matter most.
01 / thesis
The Expertise Transition · 1 Nov 2026
Software is no longer the moat.
01
Software is no longer the moat
For thirty years, the hard part of software was writing it. Code was scarce, engineers were expensive, and whoever shipped first could hold a market for a decade.
That era is ending. When code is cheap and every competitor runs on the same models, a feature can be copied in a quarter. The popular conclusion is that software margins are heading to zero.
That conclusion is half right. Software alone is no longer a moat. But every field still has moats of its own, and they are deeper than ever for the companies that know where to find them.
02
Expert work was limited by the expert
Work follows its binding constraint. For the last half century, the constraint on expert work was the expert. If you couldn't find or afford a specialist, the work didn't get done, or it got done slowly, expensively, and offshore.
Whole industries grew up around that scarcity. Outsourcing firms, agencies, and billable-hour practices all sold access to specialists by the hour. Software helped at the edges, but it sold seats to the people doing the work. It never did the work itself.
That constraint is lifting. When an agent can do the job to a professional standard, expertise stops being rationed. The work can move to wherever the need is, not wherever the specialists happen to be.
03
Retrofitting won't get us there
This transition will not be won by bolting AI onto yesterday's software. A chat panel on top of a database is not a product. A copilot in the sidebar of a legacy tool makes a specialist slightly faster. It does not do the job.
AI-native means starting from a blank page. The agent is the primary operator, and the person is the reviewer. Every data model, permission, audit trail, and screen is designed around that from day one.
The bar is the field's own standard. In work where being wrong is expensive, output has to survive a close reading by the professional who signs off on it. If the software only advises, it isn't finished.
04
Every field has its own moat
The moats that survive cheap code are specific to a field. They are different in each one, and they are rarely where an outsider would look.
Trust in the evidence. In some fields the moat is output a professional will sign: every conclusion tied to a source they can check.
The network. In others it is the people who already gather in one place, and who become more valuable to each other once they are connected.
The record and the risk. Elsewhere it is owning the system of record, holding the license, or carrying the liability that a tool vendor won't.
A generic tool captures none of these. A point solution captures one by accident. The durable company is the one that knows which moat its field rewards and builds the whole stack around it: data, agents, human review, integrations, and pricing tied to the work delivered.
02 / portfolio
5 shipped
Our products
02
AI patent agent
An AI agent for patent work: drafting, prior art, and prosecution workflow.
legal and IP
patent attorneys, prosecution teams
03
AI risk intelligence
AI financial risk intelligence that surfaces exposure and risk signals across a portfolio.
finance
individual and institutional investors
04
AI family assistant
Helps parents in high-stress co-parenting become calmer, with more time to spend on the child and less on the stressful management of co-parenting affairs.
family and co-parenting
parents, high-stress co-parents
05
AI relationship assistant
The new way to find and nurture romantic relationships.
relationships
people looking for a romantic relationship, couples
03 / how we build
How we build
supercomplicated is an AI-native company that builds AI-native products. We build them, and we run them.
01
We start with an operator, not a market map.
Moats are found by people who have done the work. They know where the job is actually won and lost, what a reviewer checks first, and what a customer would never switch away from. We open a field when someone like that wants to run it.
02
We enter where the expert is the bottleneck.
We go where an agent can do the whole job, and where needing a specialist is what keeps most people from getting it done.
03
We build the stack around the moat.
Once we know what compounds in a field, we design everything from a blank page to capture it. The agent does the work. The output meets the field's standard.
04
Shared underneath, separate on top.
Every product runs on one foundation: model access, evaluation, security review, and design. Brand, pricing, support, and roadmap belong to each product and the people it serves.
a system that compounds
Every field we enter teaches us something the next one needs. Each product sharpens how we find a moat, adds to the shared foundation, and raises the standard our agents are held to.
The first product makes the second easier. The second makes the third faster. What compounds is not the code. It is the method: turning an operator's expertise into agents that do the work to a professional standard.
Software is no longer the moat. Knowing which moat a field rewards, and building the whole stack around it, is.