Founder sues YC over idea in rejected 2018 application

Why it matters

Accelerators routinely collect detailed applications from founders without signing nondisclosure agreements. This case tests whether an application and an alleged oral confidentiality understanding can create liability when an investor later builds a related product. Four claims have survived demurrers, but those rulings establish no copying by YC.

Harmony Oswald applied to Y Combinator with a startup called Founderology on March 24th, 2018.

The application introduced Oswald as a former NBC Fear Factor contestant. RuntimeWire reviewed the episode on Prime Video: Oswald completed its opening helicopter-platform water stunt but was eliminated before the second challenge due to having the slowest time.

Founderology's YC run also ended before round two. The startup was not selected for an interview.

Three years later, YC launched Co-Founder Matching, a free service that recommends potential startup partners. Oswald discovered the product in August 2024 and sued Y Combinator Management and several affiliated funds the following May, alleging that YC used her rejected application to build it.

The sequence is straightforward: Oswald pitched a founder-matching service, YC rejected it, and YC later launched a product in the same category. Her lawsuit turns on what happened inside YC during the intervening three years.

The public record has not answered that question. It contains no email, access log, product document or witness account showing that Oswald's application reached the people who built Co-Founder Matching. Her application acknowledged that rival services already existed, and her latest complaint says Founderology had no code. YC's public product description also omits the investor-informed method Oswald identifies as her trade secret.

The case has nevertheless survived three amended complaints and several attempts to dismiss it. Oswald, a California attorney representing herself, said in April that four claims will proceed: trade-secret misappropriation, breach of implied contract, negligence and unfair competition.

No court has found that YC copied Founderology. The case remains at the pleading stage, where properly alleged facts are generally accepted as true and evidence has not yet been weighed.

The application described a familiar product category

Oswald described Founderology as an "app + platform" that would rank and suggest team members throughout a company's life cycle. She called its ranking system the I-2xD, or Investor Due Diligence, Method and selected "Artificial Intelligence" as the company's category.

The application described Founderology as a private beta with a prototype, no revenue and 20 people signed up through its landing page. In her Third Amended Complaint, filed in March 2026, Oswald says she "had not yet developed the code." She argues that the trade secret instead consisted of a method, process and business logic.

Oswald and co-founder Ginny Townsend also submitted a short application video pitching a service for finding co-founders, team members and strategic alliances. They described Founderology as being in beta testing and identified subscriptions as its business model. They did not discuss artificial intelligence, investor due diligence, the I-2xD method or a confidential matching process.

Oswald's complaint characterizes the recording as a high-level pitch that intentionally excluded confidential information. Under that account, the alleged trade secret was disclosed through the written application and a disputed oral confidentiality exchange, not the video.

The written application named LinkedIn, Founder Dating, Founder2Be and other founder networking groups as competitors. Founder2be had launched a co-founder skill-matching network in 2011, while CoFoundersLab and FounderDating merged in 2016, combining two entrepreneur-matching networks two years before Oswald applied to YC.

That history does not prevent Oswald from asserting a trade secret. California law can protect a specific method or process if it derives economic value from secrecy and its owner takes reasonable steps to keep it secret. But the distinction matters because Oswald's pleadings move between a narrow investor-informed method and the broader allegation that YC built the co-founder matching platform she invented.

Calling Founderology an AI company does not bridge that gap. Oswald acknowledges that the startup had no code, and YC's launch announcement does not describe its matching system as AI.

Founderology and YC described different matching methods

YC's July 2021 announcement says users provide their interests, location, skills and preferences. The service then displays profiles that most closely match the user's stated ideal co-founder. YC compared the experience to online dating.

Founderology proposed a different approach in the passages Oswald identifies as proprietary. The product would recommend the team an investor or startup adviser believed a company should have, even when that conflicted with the founder's preferences. The application also discussed subscriptions, corporate fees and potentially taking equity in startups formed through the service. YC offers its service for free, without taking equity simply for making a match.

YC now says its engine incorporates knowledge accumulated from successful founding teams. That language could become relevant during discovery, but its public pages do not say the service ranks people using investor due-diligence data, Oswald's I-2xD method or her proposed equity model.

Oswald's exhibits identify Catheryn Li as the lead developer who built YC's product. Li also wrote the July 2021 launch post. The released materials establish her role, but they do not show that she accessed Oswald's application, received its contents from another YC employee or implemented the I-2xD method.

Internal access records, product proposals and development histories could change that picture. None appear in the public record.

The missing link is access

Founderology's application was one of thousands YC received during the period. YC said its Winter 2019 batch alone drew more than 12,000 applications, after receiving roughly 9,200 for Summer 2018, the batch to which Founderology applied. Available figures indicate that YC received more than 32,000 additional applications across the next three cycles before Li joined as a product engineer in March 2020.

YC's current FAQ says applicants can reference past submissions, and its privacy policy says the organization uses application information to track applicants, founders, companies, industries and trends.

Those materials show that application data can persist. They do not establish how long Oswald's application was retained, what permissions Li had or whether anyone retrieved Founderology while developing Co-Founder Matching. The released record contains no database log, internal search, message or product specification connecting the application to Li's work.

The confidentiality allegation became more specific

Oswald's original complaint cited a nondisclosure agreement she signed with a contractor involved in the Founderology prototype. YC was not a party to that agreement. An article included among Oswald's exhibits also says she understood that investors such as YC generally do not sign NDAs and should not be asked to sign one. She said she relied instead on YC's culture, ethics policies and reputation for handshake deals.

YC has argued that the application contained no confidentiality designation or reference to confidentiality protections, according to a public post by Oswald quoting YC's court filing.

Oswald's handwritten notes from a March 23rd, 2018 encounter with YC partner Kat Manalac discuss Founderology's potential customers, YC's founder demographics and advice for completing the application. The notes do not record an agreement to keep Founderology confidential or a promise that YC would compensate Oswald if it later used her idea.

The original complaint likewise omitted an express confidentiality exchange. Oswald's Second and Third Amended Complaints add one. She now alleges that she told Manalac she had a valuable confidential invention and that Manalac said she understood before instructing Oswald to include the information in a YC application.

That allegation may be tested through testimony and discovery. The difference between the contemporaneous notes, the original complaint and the later amended versions will also be part of the dispute.

The patent claim was a provisional application

Oswald says she filed a "provisional patent" for "Ranking Co-Founder Suggestions Based on Data from Investors" in December 2017. The filing was a provisional patent application, which the U.S. Patent and Trademark Office does not examine on its merits. Such applications are abandoned after 12 months unless followed by the required nonprovisional filing.

Oswald says she decided against spending more than $10,000 to continue pursuing patent protection after YC rejected Founderology. She interpreted the rejection as evidence that the venture was less promising and moved on to another company.

The public exhibit package includes the USPTO receipt, cover sheet and administrative documents, but not the five-page specification describing the claimed invention. Oswald's latest complaint reserves the right to further define the trade secret after obtaining additional USPTO records and discovery.

Why four claims survived

YC's first demurrer was sustained in October 2025, with permission for Oswald to amend. In January 2026, the court allowed her trade-secret and implied-contract claims to continue. Oswald filed her Third Amended Complaint in March. On April 27th, the court overruled YC's demurrer to the negligence and unfair-competition claims, leaving four claims alive, according to Oswald's account.

A demurrer tests whether a complaint alleges a legally sufficient claim. It does not determine whether the plaintiff can prove those allegations. Once Oswald alleged an express confidentiality understanding, a specific investor-informed method and later use by YC, the court could allow the case to proceed without deciding whether the available evidence supports her account.

The rulings therefore do not establish that Founderology was a trade secret, that YC owed Oswald a duty, that YC's developers saw her application or that Co-Founder Matching uses her method. Those questions remain for the next stage of the case.

The damages theory looks decades ahead

Oswald's complaint says YC invested in more than 50 companies whose founders met through Co-Founder Matching between 2021 and September 2024. It applies a claimed 5.5% YC unicorn rate, assumes each resulting unicorn is worth at least $1 billion and projects 18.4 such companies by 2047. Applying YC's 7% standard ownership stake produces approximately $1.288 billion in projected investment value.

The model attributes the future value of those companies to the matching service and carries the observed company-formation rate forward for two decades. The complaint says precise damages will be determined through discovery. The court has not endorsed the calculation.

What the public record shows

Oswald could still obtain evidence that materially strengthens her case. Records showing that YC's product team retrieved her application, discussed Founderology or implemented the I-2xD process would provide the factual connection currently missing.

For now, the documents show that Oswald submitted a high-level proposal in an established product category, had not developed code and was rejected before the interview stage. Three years later, YC launched a free matching service that its public materials describe as using interests, skills, location and user preferences.

The documents also show that Oswald's allegations became more specific as the lawsuit was amended. The express oral confidentiality agreement does not appear in the original complaint or the handwritten notes from the conversation. The patent invoked in the pleadings was an unexamined provisional application that Oswald did not pursue.

Oswald's central theory remains testable: YC possessed her application and later built a product in the same category. The public record has not yet shown whether her pitch reached the people who built that product or whether YC used the investor-informed method she claims as a trade secret.

Investors can hear hundreds of pitches without each rejected idea attaching a lien to whatever they later fund or build. Otherwise, venture firms would need fewer partners and far more process servers. That arrangement has kept the startup ecosystem remarkably harmonious.

The lawsuit is Harmony Oswald v. Y Combinator Management LLC et al., San Francisco Superior Court case No. CGC-25-625277.

RuntimeWire reviewed the original complaint and exhibits, the Second and Third Amended Complaints, publicly indexed court rulings, Oswald's public statements, YC's product pages and contemporaneous reporting on earlier co-founder matching services.