Before João de Paula had an AI company, a $56 million funding announcement or a second Y Combinator badge, he had a law degree and a friend with an inconvenient opinion. Roberto Riccio, who would become his co-founder, told him that an early-stage startup had limited use for someone who was only a lawyer. It was rude in the productive way: specific enough to sting, useful enough to change a life.
De Paula had studied at the Universidade Federal do Rio de Janeiro and worked in law. He wanted to build, so he replaced a portion of his legal reading with programming and entrepreneurship books. By 2011, he and Riccio were working on Glio, a local discovery network where Brazilians could review restaurants, shops, cinemas and other businesses. They began with savings, a small room at a Rio university and working days that could stretch to fifteen hours. Glamour, at this stage, was mostly a chair nobody else had booked.
Glio applied to Y Combinator four times. The fourth answer was yes. In 2013, it became the first Latin American startup admitted to the program, moving its founders from Rio to Mountain View. The young company had 11,000 registered users and 160,000 monthly visits by the following year. Twenty-six investors backed it. De Paula talked about advertising, regional expansion and the clock. His sharpest observation from that period was plain: a startup’s fundamental asset is time.
Act I / The company that did not last
Seven years is a long time to earn a short lesson
Glio did not become the Latin American consumer giant its founders imagined. After roughly seven years, they closed it. De Paula has since spoken about the company without sanding off the failure. He discussed how much energy went into the wrong version of heroism, how a startup can consume its operators and how closing one leaves knowledge that cannot be displayed on a cap table.
The failure also made a useful distinction visible. Persistence had opened YC’s door, but endurance alone could not make Glio’s market cooperate. Founders adore the word grit because it can make suffering look strategic. De Paula’s later reflections are more discriminating. Time and energy are company resources. Spending both merely proves they were available.
“The most important thing we learned is that a startup’s fundamental asset is time.”João de Paula, during the Glio years
After Glio, he took six months away and then returned to startups with Matthew Watson. Together they founded Origin Financial in San Francisco in 2018. De Paula became CTO. The company built a financial platform offered through employers, and his remit grew with its product and engineering organization. In 2020, he hired Lucas Silva as an early engineering manager. Sergio Bergmann also became part of the engineering leadership orbit. Those relationships would matter later.
Origin raised a $56 million Series B in 2021. De Paula’s comments at the time sound like an operator looking past the confetti. He said the money would go toward product, technology, staff and international expansion. He also declined to treat fundraising as the achievement itself. Capital was a means to move faster. The distinction is elementary and therefore frequently ignored.
Three companies and the connective tissue
Act II / The interval
A sabbatical asks the awkward question
De Paula moved from Origin’s CTO role to an advisory position, then took a sabbatical. He allowed for the possibility that he might be finished with startups. This is the sort of freedom founders claim to want and often find alarming when it arrives. Without the next sprint, the question is no longer what must ship. It is whether the ship should leave at all.
He later described the interval through Heraclitus: the river had changed, and so had he. The rapid spread of generative AI altered what software teams could do. His own appetite to build remained. More important, an old operating frustration looked newly solvable.
At Glio and Origin, growth had made the work harder to see. Each customer, employee and feature added coordination. Leaders compensated with meetings, project tracking and requests for updates. People performed work, then performed a small pageant explaining the work. Problems could remain politely invisible until they broke something expensive.
“People who treat themselves as a work in progress have the best chances in a world that is changing so quickly.”João de Paula on adaptability
Mesmer started on January 1, 2025. De Paula took the CEO role and asked Silva to become CTO, saying publicly that Silva could do that job better than he could. Bergmann joined as a third co-founder and engineering leader. The team was less a reunion tour than a practical wager on known working relationships. De Paula later said that hiring people he already trusted removed the background worry about whether colleagues were capable, committed or moving. The external CEO work remained difficult. Inside the company, he was having fun.
Act III / Making work observable
The meeting is a symptom
Mesmer entered Y Combinator’s Spring 2025 batch, twelve years after Glio. Its first public framing was “Cursor for CTOs,” an AI chief of staff that could read evidence across code, tickets, specifications, designs and conversations. Instead of asking every participant to narrate a project, the software would assemble a live view: what shipped, what stalled, where risk was accumulating and who might unblock it.
The coordination tax Mesmer is designed to compress
At launch, Mesmer said it was working with more than ten companies collectively valued above $3 billion and had reviewed more than 70 million lines of code. Its product language has since expanded from reporting toward an operating layer for agentic engineering teams. The durable idea underneath the changing copy is observability: if leaders can see the actual work, they can remove some of the bureaucracy created to compensate for blindness.
The premise carries an appealing danger. Work is not merely the residue it leaves in GitHub, Slack or a project tracker. Judgment, trust and quiet assistance can evade instrumentation. Mesmer’s challenge is to create clarity without turning every human exchange into a management sensor. De Paula’s background makes this tension interesting. He has occupied the engineer’s chair and the executive’s chair. He knows why the dashboard is tempting and why the people represented by it may roll their eyes.
His return to YC also produced advice fit for someone carrying a well-stocked cabinet of memories. Dalton Caldwell told him not to overlearn his lessons. Experience can prevent a repeat mistake; it can also turn one arbitrary wound into a universal prohibition. A scar is evidence that something happened once, not legislation for the future.
De Paula applied that idea to fundraising. He has written about telling the same pitch twelve to fifteen times a day, investors disappearing after warm conversations and one person leaving mid-meeting upon hearing there was no revenue. During Mesmer’s round, he met another founder who had just closed and immediately questioned his own decisions. Caldwell looked at the numbers and told him to keep doing the meetings. Mesmer closed its round a few hours later.
“Fundraising messes with your head because you’re comparing yourself to people whose situation has almost nothing to do with yours. It’s a single-player game.”João de Paula, reflecting on the Mesmer round
The useful thing about coming back
By the end of 2025, de Paula said Mesmer had grown fivefold over several months and was seeing customers pull the product into their work. He named three wins: starting again, finding early signs of product-market fit and building a team he liked working with. None is a victory-lap metric. Together, they describe a founder watching for proof that the company is becoming less hypothetical.
His public definition of success offers the cleanest view of the aspiration: freedom from external obligations, with room not to be reactive. Mesmer is a corporate version of that wish. It tries to give engineering leaders enough context that they can spend less time responding to surprises and more time changing the system that produces them.
There is a neat comic loop in the lawyer becoming a coder, the coder becoming a CTO, and the CTO building software for executives. Each role teaches him why the previous one was insufficient. Law supplied structure. Code supplied agency. Management supplied a close view of coordination’s tax. A sabbatical supplied the empty space in which those lessons could finally introduce themselves.
The story remains in progress, which is where de Paula seems to prefer people. Mesmer still has to prove how much organizational bureaucracy software can safely absorb. Its founder has already proved something narrower and more portable: changing direction need not erase the road behind you. The trick is to bring forward the evidence, leave behind the costume drama and keep enough room for the next lesson to contradict the last.