The useful thing about a dead bill is that it leaves fingerprints. In Nevada, Armand Iorgulescu and his future Fed10 co-founders, Zayan Islam and Winston Wei, had spent months doing the slow civic work: building support, studying language, watching calendars, and trying to move an idea through a system designed to make haste feel suspicious. Then, 48 hours before the final vote, someone slipped a line into an unrelated amendment. Their bill died. The consultants and software paid to notice precisely that sort of movement did not notice it in time.
It was an expensive lesson in political physics. A proposal can gather endorsements, testimony, and momentum, yet one quiet sentence can still alter its orbit. In his account of the episode, Iorgulescu first assumed the team had simply been unlucky. Then the trio spoke with hundreds of policy teams. The complaints repeated: large bills from lobbyists and tracking vendors, hours lost to monitoring, and consequential changes discovered after the useful moment had passed.
The young lobbyists did what founders are forever advised to do and rarely manage with such neatness. They stopped treating the wound as private. They checked whether everyone else had the same scar.
“Some problems you can't walk away from.”Armand Iorgulescu
Before the startup vocabulary
Iorgulescu's route into software began without much software. He grew up in Las Vegas in a first-generation American family. His parents had immigrated from Romania in 1997, seeking the political freedom they had not known under communism. In written testimony to the Nevada Assembly in 2025, he described their attempt to vote in the 2008 presidential election. Election materials were not available in Romanian, and the process they wanted to join became harder to navigate because of language.
That memory became political instruction. During the 2024 election, Iorgulescu volunteered in election protection and helped limited-English-proficiency voters find assistance. The next year, while still a high-school student, he testified for Nevada Assembly Bill 367. The measure expanded language access for voters, including translated online information, interpretation help, and a language-access coordinator. It was signed into law in 2025, with its main provisions taking effect on January 1, 2026.
His testimony made the stakes plain without decorating them: a democracy may promise participation and still put its instructions out of reach. “Democracy is only as accessible as the language it is written in,” he wrote. It is the kind of sentence that belongs equally in a hearing record and a product memo. Access is not an abstraction when the missing information changes what a person can do.
By then, public service had already become more than an extracurricular label. In 2024, Iorgulescu was selected as one of Nevada's two senators to American Legion Boys Nation. He also joined Nevada's Advisory Council for Financial Literacy as the Clark County student member, appointed for a term ending in 2026. Y Combinator's profile of him says he and his friends skipped high-school classes to sit in on law-school lectures. Truancy has seldom been so bureaucratic.
Three friends, one procedural disaster
The Fed10 founding story is communal. Iorgulescu, Islam, and Wei were friends before they were executives. They worked in nonprofit advocacy and learned legislation as participants, not spectators. They knew that public policy does not move solely through speeches. It moves through version histories, committee substitutions, stakeholder calls, and language so dull it can conceal a small revolution.
That experience gave them a specific complaint about the tools on offer. Policy teams could pay consultants tens of thousands of dollars a month. They could subscribe to legislative databases. They could still be surprised. The problem was not a shortage of documents. It was the distance between a new document appearing and someone understanding what it meant for a particular organization.
Fed10 is built around shortening that distance. The company describes itself as a legislative consulting firm staffed by AI agents. Those agents monitor bills, identify threats to a business, explain the consequences, draft possible amendments, and assemble an advocacy strategy. In the founders' telling, software handles the research layer so policy professionals can spend more time on the human layer: judgment, persuasion, and relationships.
The distinction matters. A database can announce that a bill changed. A policy team needs to know whether the change is harmless, costly, useful, or quietly fatal. Fed10 is trying to sell the latter. Y Combinator's company page recounts a user facing a proposal that could have produced $10 million a year in compliance costs. Fed10 says it caught the issue, drafted an amendment, and produced an advocacy plan in seconds.
That is the commercial version of the Nevada lesson: finding out is only valuable while there is still time to act.
Harvard, briefly
Iorgulescu arrived at Harvard to study government. Fed10 was among the projects in the Harvard Kennedy School Institute of Politics Governance Lab in fall 2025. He might have continued along a recognizable path toward law, policy, or public service. Then came an email from Y Combinator partner Jared Friedman. Fed10 had entered the accelerator's Winter 2026 batch.
Iorgulescu left after one semester. Islam left Berkeley, where he had studied computer science and government. Wei left Williams, where he studied computer science and American studies. The three moved from college timetables to the compressed clock of a YC batch, building in San Francisco.
“A year ago, I didn't know what a startup was.”Iorgulescu, announcing Fed10's YC acceptance
The line is charming because startup circles are usually populated by people who claim to have incorporated during recess. Iorgulescu's advantage was different. He knew the work before he knew the vocabulary around financing it. The company emerged from committee rooms, advocacy campaigns, and the stomach-drop of discovering a decisive change too late.
That sequence offers a useful principle. Begin with the job. Do it by hand. Stay close enough to notice where competent people repeatedly lose time or money. Only then give the problem a product name. Fed10's founders did not browse a market map and select “government affairs.” They had already been living in its least glamorous tabs.
The part software should not replace
There is a minor paradox in automating lobbying research. Politics is stubbornly human. A vote can turn on trust, timing, constituency pressure, or a conversation no model observed. Fed10's pitch does not erase that reality. It depends on it. If machines absorb the searching, comparing, and first-draft analysis, humans can spend more of their hours maintaining the relationships that move legislation.
Iorgulescu's own story argues for keeping people in the frame. AB367 was not born from a synthetic benchmark. It was connected to his parents' encounter with the practical limits of voting access, then sharpened by his experience helping voters in 2024. The resulting testimony had force because a general policy failure had entered one family's kitchen-table history.
The same goes for the founding team. On paper, their profiles make a startling sequence: bills before ballots, state appointment, Harvard, Berkeley, Williams, YC. In life, the connective tissue is friendship. They attended law-school lectures together, did advocacy together, lost a bill together, and built the response together. The company is not a lone prodigy's thunderbolt. It is a group project that refused to end when the grade was posted.
Radiohead B-sides and other evidence of life
A founder can be flattened quickly by a batch label. Iorgulescu's tiny personal website provides a welcome correction. Its interests page contains four items: Bob Dylan, Linsanity, David Foster Wallace's cruise-ship essay “Shipping Out,” and Radiohead B-sides. This is both specific and gloriously useless to a customer-acquisition funnel.
It also suggests a taste for cult runs, wordy observation, and material beyond the obvious single. Perhaps there is a line from legislative amendments to B-sides: consequential things often hide away from the headline. Perhaps that is too neat. A person is allowed to like “Polyethylene” without turning it into a management philosophy.
Fed10 is still an early company. Y Combinator lists it as active, founded in 2025, with the three co-founders building in San Francisco. Its ambition is large because the document stream is large: legislation and regulation across jurisdictions, each change potentially material to someone. The work ahead is to prove that speed can coexist with accuracy and that automated analysis can earn trust in a field where phrasing is consequence.
Iorgulescu's story so far is less a victory lap than a change of venue. The committee-room instincts remain. Read closely. Ask who is affected. Notice the late line. Make the information usable before the clock runs out. Politics taught him that fine print has a body count measured in missed rights, unwanted costs, and lost bills. Fed10 is his attempt to make sure it also has an alarm.