Three years, three different problems. What a founder's record actually looks like by the time anyone files.

What each year contributes, and what it does not. The regulatory basis for every row is in the sources at the foot of this page.
Year	What the record gains
Year one	A field, named narrowly. Nothing else. No criterion is met and no filing is possible. The one decision that changes how every later document reads is which field the person will be recognised in, because it sets the comparison group for everything afterwards.
Year two	The first evidence that does not belong to the company. Independent judging, third-party technical writing, membership in bodies that admit on judged achievement, and awards received by the person rather than by the employer. This is where criteria start being met on paper.
Year three	The two things the final merits determination actually turns on: documented use of the work by other people in the field, and recognition from bodies qualified to judge the field rather than from media companies or accelerators.
The year it is filed	Assembly, and almost nothing else. Letters, translations, certified records, a coherent field definition, a plan for continuing in the area. A filing year that generates new evidence is a filing year that arrived late.
When it goes wrong	A denial or a Request for Evidence usually traces back to year one rather than to the filing. The field was too broad to have a top, or the recognition being claimed was the employer's and not the person's.

The short version

Year one of a founder's record produces no evidence of anything, and that is the part nobody says out loud. The first year is spent choosing a field narrowly enough that a specific body of people can form a judgement about the top of it. That choice, not the product, decides how every later document is read, because the field sets the comparison group for every criterion that follows. A founder with a seed round and thirty pieces of launch press has met one criterion weakly and failed the rest. The same founder three years later has an entirely different problem, and a much better one.

Years two and three are where criteria are genuinely met, and they are met by work that was never done for a petition. Peer review for a conference, an open-source project that other people depend on, a standards committee, a selection panel, a technical talk accepted by people who know the subject — none of that is filing activity. It is ordinary participation in a field, and it is the only reliable source of the third-party recognition the criteria are built around. There is no shortcut that produces it in a quarter, because the thing being measured is other people's independent judgement, and those people have to have a reason to form one.

The filing year itself contributes assembly. USCIS assesses the case in two steps: first, determine by a preponderance of the evidence which evidence objectively meets the description of each regulatory criterion; second, evaluate the whole petition together in a final merits determination, against the standard of whether the person is one of that small percentage who have risen to the very top of the field of endeavour. Passing step one is expressly not eligibility. The guidance states that satisfying step one alone does not establish that the person in fact meets the requirements for the classification. Founders who file late almost always arrive with step one and no step two.

Preponderance is a low bar applied one fact at a time and a hard bar applied to a record. A person may be very young or early in a career and still show sustained acclaim; there is no definitive time frame on what constitutes sustained, and the guidance says the term does not imply an age limit. What it does imply is maintenance. An officer determines whether the person continues to maintain a comparable level of acclaim in the field since the recognition was originally afforded, and is told plainly that a person may have achieved national or international acclaim in the past but then failed to maintain a comparable level of acclaim thereafter. What matters is not the peak. It is whether the level is still held.

The asymmetry that decides more founder cases than any other is this: coverage of the company is not coverage of the person. USCIS directs that published material should be about the person, relating to the person's work in the field, and not just about the person's employer and the employer's work or about another organisation and that organisation's work. Marketing material created for the purpose of selling the person's products or promoting the person's services is not generally considered published material about the person. A founder's press page is, structurally, half an asset and half a liability.

The observed pattern, stated as plainly as it can be: the record that succeeds is the one built before anyone decided to file. The founder who starts treating the criteria as a description of a career rather than a checklist for a filing spends three years producing exactly the documents an officer will look for, and then spends a quarter assembling them. The founder who starts in the filing year produces a great deal of material about a company and very little about themselves.

Year one: the field, defined before the work starts

The regulation places extraordinary ability in the sciences, arts, education, business or athletics, and nothing about that list is the founder's problem. The founder's problem is the field claim inside it. A field has to be narrow enough that a specific community can form a judgement about who is at the top of it, and broad enough that the person is not alone in it. Distributed-systems reliability for financial infrastructure can be led. Technology cannot, because no body of people is qualified to say who is at the top of technology. Software engineering as a whole is a second bad answer for the same reason.

This is a decision with no visible consequences for two years, which is exactly why it is usually made by accident. A founder who describes themselves as a generalist, or who lets the company's category become their category — a fintech founder, a marketplace founder — has chosen a field too broad to have a top and too commercial to be judged by experts. The field does not need to be obscure. It needs to be a body of specialised knowledge that exists independently of the company.

Year one's real output is position rather than documents, and position is set by which institutions the person joins. A membership qualifies under criterion two only where the association requires outstanding achievement of its members, as judged by recognised national or international experts. General membership does not qualify, membership based on years of experience does not qualify, and membership obtained by paying a fee or subscribing to publications does not qualify. The instruction to join the body that judges the work rather than the body that markets it sounds like career advice. It is also, three years later, the difference between a criterion met and a criterion absent.

The same logic applies to the first awards worth bothering with. Recognition from an accelerator, an incubator, a large customer or a media company carries no weight in a comparison group made of people who do the work, because the awarding organisation is not qualified to judge the field. A selection run by a field body — a technical society, a peer-reviewed venue, a standards group — is a different instrument entirely, even when the prize is smaller and less photogenic. Year one is the only point at which that trade can be made cheaply.

Year two: the first evidence that is not the company's

The result of year two, done properly, is a record that has moved from company-shaped to person-shaped. The documents that do that work are unglamorous: review invitations accepted, a technical piece placed with people who care about the subject, a committee seat, a citation trail. None of it is impressive in a pitch deck. All of it is legible to an officer applying a criterion written for scientists, and legibility is the whole game.

The most common year-two error is volume. A founder who publishes, speaks and posts constantly, all of it about the company's market position, produces a large archive of material that is about the employer and the employer's work. The guidance excludes exactly that. Five documents about the person beat five hundred about the company, and the arithmetic is not close.

The easy criteria arrive first, and they mislead

Year two reliably produces two things. The guidance states that participating in the judging of the work of others in the same or an allied field of specialisation alone, regardless of the circumstances, should satisfy the regulatory criteria in the first step of the analysis. It states the same for publishing scholarly articles in professional or major trade publications or other major media alone, regardless of the calibre. Both are accordingly easy to evidence and, on their own, weak. The second step requires the officer to evaluate whether that judging, or those publications, were indicative of the person being one of the small percentage who have risen to the very top of the field. A founder who collects both in year two has cleared a threshold and has not yet made an argument.

Awards have to be the person's, not the employer's

The guidance is specific that the awards criterion concerns the person's receipt of the awards or prizes, as opposed to the employer's receipt. Company-of-the-year lists, founder-of-the-year lists run by magazines, and prizes awarded to the product are all, on their face, the employer's or the product's receipt. What counts is an award for excellence in the field of endeavour received by the person, and the awarding body's standing does most of the work. Nothing precludes relying on a team award provided the person is one of the recipients, which is useful for founders who win things jointly.

The one thing worth starting in year two because it cannot be hurried

Criterion five asks for original contributions of major significance in the field. This is the criterion that is genuinely hard and genuinely slow, and it is the one founders almost universally leave until the end. The guidance points to the kinds of evidence that are probative: research that has provoked widespread commentary on its importance from others working in the field, and documentation that it has been highly cited relative to others' work in that field. Both of those sentences describe a process that takes other people years to complete on your behalf. No amount of writing about the work substitutes for it, because the evidence is other people's behaviour.

Year three: the two criteria that carry the case

By year three the record can support the criteria that do real work at the final merits step: documented use of the work by others in the field, recognition from bodies qualified to judge it, an elevated remuneration figure relative to others in the same field, and a leading or critical role at an organisation with a distinguished reputation. A founder's role claim is usually strong, and it is usually argued badly — the role has to be leading or critical to the organisation, and the organisation's distinguished reputation has to be documented independently of the founder asserting it. Investor logos and a valuation are not documentation of a distinguished reputation.

Year three is also the year the maintenance question becomes live. Because sustained is read as maintained, a record that peaks in year one and goes quiet is worse than a record that rises steadily into the filing. A founder who won a genuinely competitive award at twenty-four and has done nothing judged since will be asked what level of acclaim has been held since. The answer is not the award. The answer is whatever has happened in the field in the years between, which is why the interval is not recoverable later.

The third thing year three decides is whether the person still works in the area of their claimed ability. The regulation requires that the person intends to continue work in the area of expertise, and sets out that the evidence may consist of letters from prospective employers, evidence of prearranged commitments such as contracts, or a statement from the person detailing how they intend to continue working in the field. This catches founders specifically, because the successful founder is often promoted out of the discipline that produced their record. Moving from building to fundraising, or from an operating role into general management, changes the answer to a requirement that has to be satisfied at filing time, not at the time the work was done.

None of this is a reason to delay building the company. It is a reason to notice that the company's success and the person's record are two separate assets that happen to be produced by the same three years of work. The ones who file easily are the ones who treated the second asset as real from the beginning, without ever saying the word petition.

The filing year, and what it is actually for

The filing year is for assembly: writing the field definition so it survives being read literally, obtaining letters from people who can speak to the work rather than people with impressive titles, translating and certifying records, and drafting the statement about continuing in the area. Every one of those tasks is a fortnight. Together they are a quarter. What the filing year cannot do is create evidence that other people never generated, and the reason is the two-step analysis rather than any procedural rule. A petition can be strong at step one and denied at step two, and step two is the part that cannot be drafted in a quarter.

There is a parallel route worth naming for timing rather than for outcome. An O-1A requires at least three different types of documentation corresponding to those listed in the regulations, or comparable evidence in certain circumstances, and the evidence as a whole must demonstrate that the person meets the relevant standards. It is filed by a US employer, a US agent, or a foreign employer through a US agent, and it cannot be filed more than one year before the services are actually needed. For a founder who has their own company, the agent route is available, which means the evidence base and the immigrant case can be developed on the same schedule. Founders commonly discover that the O-1 and the EB-1A are the same record wearing two forms.

Processing is the last thing to plan and the first thing people worry about. Form I-140 is eligible for premium processing, which for most classifications carries a guaranteed action within 15 business days. The important detail for a founder planning around a start date is that the clock stops and resets when USCIS issues a request for evidence or a notice of intent to deny, and a new period begins when the response is received. A filing year with an RFE is a filing year plus a response cycle, and the response cycle is where a record assembled in a hurry is exposed.

For nearly every reading of this, the conclusion is the same one the pattern produces. There is no filing strategy that improves a thin record, and there is no filing year in which the field, the maintenance and the third-party use can be manufactured. What a founder controls is whether the next three years produce material about themselves or material about the company, and that choice is available in year one for free and in year three only at the cost of starting the clock again.

Questions this raises

How long does a founder's extraordinary ability record realistically take to build?

Two to four years of field participation is the observed pattern, because two criteria depend on other people's judgement and behaviour rather than on documents the applicant produces. Judging and authorship can be assembled within a year. Documented use of the work by others in the field cannot be compressed, and founders most often lack it.

Can a founder file in year one if the company is doing well?

Filing is possible at any time, but a year-one record typically meets one or two criteria weakly and fails the final merits step, which asks whether the person is among the small percentage at the very top of the field. A strong company is evidence about the company, and the guidance excludes material about the employer rather than the person.

Does press coverage of my startup count as published material about me?

Generally no. USCIS directs that the material be about the person and their work in the field, not just about the employer and the employer's work. Content created to sell the person's products or promote their services is not generally considered published material about the person, which excludes most launch coverage and almost all paid placements.

What is the most common year-one mistake founders make?

Defining the field too broadly, or letting the company's market category define it. A field has to be something a body of people can judge the top of, so technology and software engineering fail as fields while a specific technical speciality can be led. Field definition sets the comparison group for every criterion and it is hardest to change later.

Does a gap in the record hurt if my early recognition was strong?

Yes. Sustained acclaim must be maintained, and USCIS directs officers to determine whether the person continues to maintain a comparable level of acclaim since it was originally afforded, noting that a person may have achieved acclaim in the past but failed to maintain a comparable level thereafter. The peak matters less than whether the level is still held.