YC Startup Tracker

In today's brief, Y Combinator finds itself at the center of two stories that underscore the growing pains of AI commercialization: while one portfolio company, Corgi, faces a PR crisis over alleged code theft via 'vibe coding,' the broader P26 batch reveals a strategic pivot from building new models to creating the infrastructure needed to make AI agents trustworthy at scale.

Together, these developments highlight how the AI startup ecosystem is grappling with both the ethical boundaries of AI-assisted development and the practical challenges of operationalizing agents.

Tracking: Y Combinator

Geography: Silicon Valley, San Francisco, United States

1. YC-backed Corgi denies stealing open-source code after viral accusation

Y Combinator-backed insurance startup Corgi denied allegations from Papermark co-founder Marc Seitz that it stole open-source software for its new Dataroom product.

Corgi CEO Nico Laqua admitted the company used "vibe coding" that replicated Papermark's visual features and language, but said no code was copied. The offending elements have already been changed.

Corgi has sent a cease-and-desist letter demanding Seitz take down his tweet, adding to a pattern of aggressive legal tactics that included a prior dispute with competitor Matcha in May.

The controversy raises questions about whether copying functionality via vibe coding — without identical source code — constitutes ethical or legal infringement, especially as AI tools make replication trivial.

Corgi's reputation for litigious behavior, including lawsuits against former employees, compounds the damage. The two-year-old startup also runs a 24-hour coffee shop and plans to expand that business.

Key facts:

- Papermark co-founder Marc Seitz accused Corgi of stealing its open-source software on X.

- Corgi CEO Nico Laqua admitted vibe coding led to replica features and identical language.

- Corgi sent a cease-and-desist letter to Seitz demanding removal of his accusation tweet.

- Corgi previously faced accusations of bullying from competitor Matcha in May.

- The two-year-old startup has sued multiple former employees and is known as litigious.

Why it matters: This incident tests how intellectual property law applies in an era of AI-assisted vibe coding, where look and feel can be replicated without copying source code directly.

It puts Y Combinator in an uncomfortable position, as two of its alums — Corgi and Papermark — clash publicly, and follows PearAI's similar cloning controversy.

For founders, the clear lesson is that vibe coding can create legal exposure even without code theft.

For investors, Corgi's escalating pattern of litigation — against competitors, former employees, and now a critic — may signal governance risks that could affect future fundraising and partner relationships.

TechCrunch

2. YC's latest batch shifts from AI models to agent infrastructure

Y Combinator's P26 batch, analyzed by Extruct AI, includes roughly 205 companies. The dominant theme is no longer building new AI models but creating the practical infrastructure and business models to make AI agents reliable at scale.

Key areas include trust and insurance layers for agents, data sourcing marketplaces, vertical-specific agents, AI-powered service businesses, and prediction market tools.

A notable outlier is Polsia, a solo-founder company with nine AI agents that raised $30 million at a $250 million valuation. Hardware remains about 10% of the batch, focused on defense and attritable systems.

Key facts:

- YC's P26 batch includes roughly 205 companies.

- Luel raised a $31.2M seed, one of the largest in YC history.

- Polsia raised $30M at $250M valuation with zero employees.

- YC passed on Mercor in 2023; it is now valued at ~$10B.

- Prediction market tools grew from 1 to 5 companies in this batch.

Why it matters: The batch signals a maturation of AI investing: the focus is leaving model makers and entering the messy operational layer. Startups building trust, data brokerage, and vertical service models may capture more value than generic AI tools.

Traditional SaaS models in verticals face disruption as AI-run service businesses demonstrate higher willingness-to-pay. The rise of prediction market infrastructure suggests a growing bet on information markets.

Defense hardware startups could reshape drone economics. Watch for oversaturation in data sourcing and whether service models scale effectively.

Quasa