Adelaida
Diaz-Roa

I have started eight companies. Two of them sold. Two of them are running right now. The other four taught me most of what I actually know.

Bogotá, sometimes New York Español / English Success Magazine 30 Under 30 Shark Tank at 18

How it started

I read about a small business in the Dallas Morning News and wrote to the owner asking to work for free. I knew a shop that size could not pay me. He said yes and put me on the Facebook account.

By the end I was rebuilding the website, hiring, managing inventory, and handling customers. That company became my first exit, and the pattern held: every useful thing I know came from being handed a job slightly beyond what I had earned, then figuring it out in public.

Everything below is the same story repeating at bigger stakes. Including the parts that did not work.

"I wrote him about wanting to help him out as a volunteer, because I knew as a small business he wouldn't be able to pay me." From my journal, on starting at Villy Customs

The whole record

Eight companies, two exits,
and four that taught me more.

Most founder pages show the wins. The four that did not exit are the reason the two that did were possible, so they get the same room here.

Villy Customs

Exit
2013 – 2017

Custom bicycles, out of Dallas. I started as an unpaid volunteer and ended up running most sides of the business.

Shark Tank at 18, in front of Mark Cuban and Barbara Corcoran. I worked under the founder for a percentage, and he kept his word about it. I left on principle over how employees were being treated. The company sold later and I was paid out.

You can be right about the money and still be wrong to stay. Integrity is the thing you cannot earn back.

Pawliday Inn

Exit
2014 – 2019

A pet resort and spa in Dallas, built with my mom. We became the top rated dog hotel in the city.

A national brand wanted into the Dallas market and bought us because we already owned it. Before that, my mom offered me a third of the business. I asked for close to half and was effectively fired. Both of those things are true about the same company.

Family and equity are two different conversations. Having one does not let you skip the other.

Ruffit

No exit
2015 – 2016

A patented pet carrier you wore. I met the inventor while handing out cards for Pawliday.

The Today Show picked it up after someone spotted the carrier on the street. British Vogue called. I wrote the Shark Tank pitch, filled out the packet, and booked the flight. I was promised ten percent and never signed for it, so when it mattered I had nothing.

Get it signed. A good relationship is not a contract, and the moment you need one is exactly when the other person stops being reasonable.

Nomo Fomo

No exit
2016 – 2019

An app about getting people into real plans with real people. Launched September 2018.

Two years of building for something scoped at three months. I was happiest in the part where I was learning fastest, and when that flattened out I lost the thread before we ever found traction. I wrote it down in my journal at the time, which is the only reason I recognized it later.

Companies die of founder misalignment far more often than they die of market rejection. Watch your own state like it is a metric.

Syzygy / Claim Deck

No exit
2019 – 2020

A founding team I had real hope for.

Communication broke, I was cut out of team events, and the contract for my equity never appeared. They used the work and planned to take what they needed. Reading about gaslighting as an actual named pattern is what let me see it while it was happening rather than years after.

Walking away from something toxic is self-respect, not failure. Also, again, get it signed. Some lessons bill you twice.

Robin House

Stalled
2023 – 2025

Fractionalized real estate for the Colombian market. We raised twelve to thirteen thousand dollars in a single day.

This is the one where I did everything right on paper. Contracts in place from day one, a real values framework, a strong culture. Then my co-founder burned out, a bad actor made everything harder, and neither of us could go fully in because we were both split across other ventures.

Splitting attention across three companies means none of them get a founder. Good structure does not survive a part-time founder.

Onde

Running now
2025 – present

Event planning for LatAm corporates. Venue sourcing and the software that runs the day.

Anchored with a real client from month one. Every lesson above is built into how it is structured, which is the only reason I am willing to run two things at once now.

Mycelium AI

Running now
2026 – present

Company memory that makes AI agents useful in production instead of impressive in demos.

Open substrate anyone can take, a paid runtime for teams who want it run properly. I use it every day on my own work before I sell it to anyone.

What I am doing now

Two companies, both on purpose.

Robin House taught me what happens when a founder is split three ways. These two are the exception I made deliberately, because one feeds the other.

Founder and CEO

Mycelium AI

Most companies have impressive AI in the demo and useless AI in production, because the model has no memory of how the company actually works. Mycelium is that memory. The pattern is open source. The runtime is the product.

Co-founder and CEO

Onde

Corporate event planning across Latin America, from sourcing the venue to running the day. Built for the teams who currently do it in spreadsheets and WhatsApp threads.

I also write

Two series, both in public.

The High-Rise

About moving between emotional floors, written from inside it rather than from above it. The operating idea is that friction is not the same thing as suffering.

After the Shock

About AI disruption and what a dignity floor looks like, from Bogotá rather than from San Francisco. Optimistic, and specific about why.

Working together

I build AI operating systems
for people who run things.

Through Mycelium AI. It starts with a deep dive into how you actually work, and if it does not show at least ten hours a month coming back to you, you do not pay.

For one person

Your own install. Your files, your voice, your decisions, all of it retrievable.

For a team

One shared brain across the company, so nobody re-answers a question that was already settled.

Ongoing

Fractional AI officer. Strategy and the system, month to month.

Start a conversation