Two Stanford PhD students entered an AI film festival with a $10,000 grand prize. They didn't win anything. So they quit their PhDs and built the thing that beat everyone who did. The company is Pika, founded in April 2023 by Demi Guo and Chenlin Meng. Guo studied math at Harvard, then computer science at Stanford, with internships at Facebook AI Research, Google Brain, Microsoft Bing, and Quora along the way. Meng, her PhD classmate, wanted to be an animator before she became her co-founder and CTO. Meng also co-authored the DDIM paper, a foundational method now used inside DALL-E 2 and Stable Diffusion. The rejection stung specifically because the tools available were bad. Existing AI video generators were clunky and hard to use, and Guo and Meng were convinced they could build something better, both the underlying model and the interface wrapped around it. Nat Friedman, the former GitHub CEO turned solo investor, saw an early demo built on a single GPU and put in the company's first check. They launched publicly focused narrowly on one thing: anime-style video generation, simple enough that, as Guo put it, "people like me and Chenlin," not professionals, could make something good on the first try. It compounded fast. A $55 million round in November 2023, backed by Lightspeed Venture Partners and a wave of angels. An $80 million round roughly a year later, led by Spark Capital, at a $700 million valuation, aimed squarely at taking on Sora, Veo, and RunwayML. By mid-2026, total funding had reached $135 million at a $470 million valuation, alongside a consumer iPhone app, an Adobe Firefly integration, an official MCP server, and Pika 2.5, a model generating 1080p video with native sound in under 90 seconds. Then, eight days ago, Pika rebuilt itself entirely. The new version drops the idea of being one video model and becomes something closer to an operating system for creative AI: video, image, and audio tools built on top of every major model worth using, Seedance, MiniMax, Wan, GPT Image, Nano Banana, Grok Imagine, ElevenLabs, and Pika's own models, all inside one interface, with a developer-facing API club offering access to all of them at once. Guo's read on the competition has never wavered. Everyone else is racing to build the best model. She's betting the company that actually wins is the one that makes creativity itself feel effortless, regardless of whose model is doing the work underneath.