Code Red

“Code Red” – a new cartoon that illustrates how OpenAI is under real pressure.

Sam Altman pulled the fire alarm last week. A “code red” is the moment in a business when the nice-to-haves get pushed off the table and everything funnels into survival priorities. For OpenAI, it means the roadmap gets rewritten around the three things that Sam thinks are the company’s core differentiation. This begins with personalization that actually remembers you, a huge jump in speed so the product feels instantaneous and deeper system-level integrations that push the model into workflows instead of keeping it trapped in a chat window. As a result, everything else gets slowed or starved, including Sora’s march toward becoming a universal video engine.

This isn’t new in tech. Meta announced a code red in 2018 when TikTok started draining culture out of Instagram. Microsoft did it in 2014 when it finally admitted mobile had crushed the Windows-first worldview. IBM basically rebooted itself in 1993. These moments aren’t necessarily heroic, they’re mechanical to avoid a perceived existential threat.

OpenAI is at this stage for a couple of reasons. Google just shipped Gemini 3.0, a step-change in model performance, which breaks the “OpenAI is always slightly ahead” assumption. Also, OpenAI’s user base hovering around 800M DAUs for months shows the ceiling of the current product. At the same time, Anthropic is quietly eating the profitable end of the market. Enterprises pay real money, and they renew. It’s why Anthropic is expected to hit profitability before OpenAI, even with a smaller user footprint.

Then there’s the structural problem. Google has ads, Meta has social, Apple has hardware, and Anthropic is building a steady enterprise business. OpenAI is still leaning on a $20 subscription and a belief that it’s the center of the AI universe. As Andy Grove liked to remind anyone who would listen, “only the paranoid survive.” OpenAI’s code red is a fight to hold its narrative together, and the consequences of that pressure will surface fast, whether through abrupt product pivots or moves none of us expected.