Cosmin Nicolaescu Has a Problem with Getting Comfortable

In October 2014, Cosmin Nicolaescu gave himself a peculiar assignment: spend a year in San Francisco without going back to the same restaurant. He had recently moved from Seattle, where familiar places had become a little too familiar. A dependable dinner was pleasant. It also meant missing whatever was cooking around the corner.

He wrote rules. Work lunches were exempt. Delivery was exempt. Brunch and dinner at the same establishment counted separately. Friends could celebrate birthdays wherever they pleased; he would spare them a lecture about his experiment. Even an adventure in eating required a specification.

The proposed year of dinners offers an unusually useful introduction to the engineer who would become Brex’s chief technology officer and later co-found Accrual. Nicolaescu likes to understand a system, make it work, then find something he has yet to learn. Comfort has a way of becoming an engineering problem.

A new country, then a server room

He was born in Bucharest. His route through North America included an exchange-student stay in Deatsville, Alabama, university in Philadelphia, and periods in Vancouver, Seattle, San Francisco and Los Angeles. These moves gave him repeated practice at arriving somewhere without the accumulated advantages of knowing how everything worked.

At Drexel University, where he studied computer science from 2003 to 2008, the work became practical early. He administered systems for the computer science department. The job included moving servers to a new storage setup and establishing automated monitoring. Before there were financial platforms to expand, there were departmental machines to keep running.

Drexel also provided a route into Microsoft through internships and then a full-time job after graduation. His degree was a beginning with several doors attached: paid technical work, a large software company and, eventually, a different country’s technology industry.

At Microsoft, he worked on early Azure and Office 365 systems and the Microsoft Commerce Platform. Payments became an interest there. He was learning the machinery underneath a transaction: the infrastructure required before a customer can do something that looks simple on a screen.

The first-day photograph

Nicolaescu joined Stripe in July 2014. His account of that first week spends surprisingly little time on grand strategy. It dwells on the laptop he could choose, the accounts already configured and the permissions he could obtain immediately. Someone had thought about the small impediments that make a newcomer useless.

His first-day exercise was to publish his photograph on Stripe’s website. A modest task, but it carried him through the development process: update the repositories, inspect the code and ship a change. The photograph was also an introduction to how the company worked. A face on a website had become an onboarding curriculum.

He noticed searchable conversations, shared notes and information that employees could reach without elaborate permission delays. In his writing, productivity was something you could damage with an ordinary administrative decision. The equipment order and the access request deserved attention alongside the architecture diagram.

His responsibilities grew into financial infrastructure, international engineering teams and Stripe Terminal. He helped establish engineering operations in Seattle, Dublin and Singapore and built Terminal’s engineering and product teams. The job brought together software, payments and the less tidy business of helping people build together across locations.

Learning to leave well

Brex announced his appointment as vice president of engineering in January 2019. He joined a company of roughly 40 people and later served as CTO as it grew beyond 1,000 employees. He had already seen Stripe grow from about 100 employees to more than 1,400. At Brex, he would encounter another version of the expanding-company puzzle.

In a 2023 conversation with Plaid CTO Jean-Denis Greze, Nicolaescu described the changing work of engineering leadership: more products, different kinds of expertise and teams that needed to evolve with the business. The organization chart could not remain a souvenir of an earlier, smaller company.

By late 2023, he wanted to build a business himself. He had been discussing his career with Brex’s founders on long walks, revisiting what interested him and where he wanted to grow. Those conversations made room for a planned departure. They agreed to promote James and hand over his responsibilities during the first half of 2024.

He treated the handover as a problem worth documenting. His involvement fell month by month; he left meetings and channels so his successor could lead without the previous CTO hovering nearby. He also acknowledged a mistake from his Stripe departure: he wished he had been more transparent earlier. Experience had given him something more useful than an immaculate career story: a correction to apply next time.

Before founding, he wrote down the questions that his experience had not settled. Could a pragmatic operator create a vision of his own? Would he remain committed for five or ten years? Becoming a founder would reduce the freedom to move on when a different opportunity appeared. He understood that commitment before Accrual had a public launch or a product announcement. The unfamiliar work was part of the attraction.

Homework for two former colleagues

His co-founder, Siddarth Chandrasekaran, was a friend from Stripe. They shared an interest in financial infrastructure and eventually returned to building together. General Catalyst backed their first two rounds; founders of Stripe and Brex were among their early supporters.

To learn accounting, the pair enrolled in community-college courses aimed at qualifying for the CPA exam. They also used their investors’ personal tax returns as a test case. It is a pleasing reversal of the usual fundraising ritual: the founders left with homework involving the investors’ finances.

Accrual launched publicly in February 2026 with $75 million raised. The initial product tackled individual tax returns, organizing documents and preparing drafts for accountants to review. It linked the work back to supporting material and connected with existing tax engines. The professional still made the decisions; the established tax system handled authoritative calculations and filing.

The choice of work suited a career spent near money’s plumbing. Tax preparation brings documents, changing information and professional review into one deadline-driven process. A missing item cannot be charmed into existence. A figure has to connect to evidence. The accountant needs to inspect how the draft was assembled.

The case for staying small

“Smaller teams are better,” Nicolaescu writes among his personal operating beliefs. He associates them with faster decisions, less overhead and more enjoyment. He is particularly wary of the consequences of hiring too many people. After years inside expanding organizations, he has a clear preference about the starting conditions.

“Smaller teams are better.”

Cosmin Nicolaescu · Personal operating beliefs

In a July 2026 interview, he described Accrual as a 21-person company. Employees shared the same salary, while he and Chandrasekaran took less. Roles were loose, work crossed functional boundaries, and the remote team spent time together in person. People used AI tools throughout the company. These were choices for that stage of the business, with experienced colleagues expected to operate independently.

There is continuity here with the new Stripe employee inspecting his laptop setup. He is still interested in how much useful work a person can do once obstacles are removed. Now he can make those choices himself, including choices about compensation and the distance between the person selling software and the person building it.

His public reflections on management ask whether a leader actually changes a project’s direction. A functioning meeting schedule is an insufficient answer. He has argued for meetings organized around producing something and for deciding carefully which problems deserve executive attention. A manager has to contribute more than a calendar invitation.

When the documents arrive

The 2026 season supplied a test beyond company philosophy. Accrual’s Armanino case study describes deployment across six offices, with thousands of returns and a median draft time of 15 minutes. Those are reported product results, specific to the work described. A draft still has to travel through review.

The operational details are more revealing than a stopwatch. Offices used local champions. Training took place in person. Client data was reconciled before the season. The rollout required people to change how they worked, as well as software that could process their documents. Nicolaescu’s organizational interests had followed him into his customers’ offices.

In September, Accrual announced an agreement to acquire Puzzle’s accounting-firm technology and business, with founder Sasha Orloff and the relevant team expected to join after closing. The plan extended into recurring client accounting work. Broader availability of the combined ledger and close capabilities was planned for the end of 2026.

Accrual had used Puzzle for its own books, giving the proposed combination an everyday familiarity. Its larger ambition reaches across tax, audit, client accounting and advisory: carry the client record and the firm’s working methods into more engagements. The same information should remain useful after the return is finished.

A more useful kind of spare time

In October 2026, Nicolaescu wrote about the first Accrual Exchange, where firm leaders discussed what to do with capacity created by AI. He returned to practical questions: how juniors develop judgment, how leaders learn the tools and whether saved time becomes a service that clients value. Faster preparation leaves plenty of decisions for the people running the firm.

His own interests extend to helping other people learn. He has described coaching founders and executives, making angel investments and hoping eventually to teach university classes. The systems administrator who learned through work has retained an interest in handing experience along.

Years before Accrual, when he restarted his blog, he explained that writing helped him organize thoughts and retain details. He had noticed that familiar work was giving him fewer new problems. So he bought a notebook and a pen, resumed side projects and started learning outside the software tools he used at his day job.

The restaurant experiment belonged to that same period. It came with sensible exceptions, a public commitment and the possibility of a disappointing meal. Nicolaescu had found a way to give curiosity a schedule. A decade later, two experienced engineers were back in accounting classes, learning enough to ask better questions.