OpenAI Calls a ‘Code Red’ + Which Model Should I Use? + The Hard Fork Review of Slop

Episode Summary

Executive Summary: In this episode of Hard Fork, hosts Kevin Roose and Casey Newton discuss OpenAI's 'code red' status due to competitive pressure from Google's Gemini 3 and Anthropic's Claude Opus 4.5. They analyze the AI landscape, including model improvements, strategic shifts, and the rise of AI-generated content ('slop'). The episode also features a segment reviewing various examples of AI slop, from fake holiday markets to AI-generated recipes and ads.

Main Topics: OpenAI's Code Red and Competitive Pressure (Priority: 5/5): OpenAI declared a code red due to concerns about ChatGPT user growth and competition from Google's Gemini 3 and Anthropic's Claude Opus 4.5. Sam Altman's memo indicated a shift in resources to improve ChatGPT, delaying other projects like ads and AI agents. Google Gemini 3 Capabilities (Priority: 4/5): Gemini 3 is praised for its speed and reliability, with 650 million monthly users. It is seen as a strong competitor to ChatGPT, leveraging Google's distribution advantage. Anthropic Claude Opus 4.5 and the 'Soul Document' (Priority: 4/5): Claude Opus 4.5 is highlighted for its empathetic and human-like text generation, with a unique 'soul document' embedded in its weights. Anthropic focuses on enterprise use and avoids consumer engagement optimization. AI Model Usage and Recommendations (Priority: 3/5): The hosts discuss practical advice for choosing AI models, emphasizing experimentation and the rapid improvement of models. They note that AI is becoming 'higher resolution' in its outputs. AI Slop and Its Cultural Impact (Priority: 3/5): The 'Hard Fork Review of Slop' segment examines various AI-generated content, including fake holiday markets, AI recipes, educational songs, unauthorized ads, and viral fake game videos. The hosts differentiate between harmful and benign slop. Industry Moves and Strategic Shifts (Priority: 2/5): Notable departures include Jan Lacun from Meta and John Andrea from Apple, signaling strategic shifts in AI development. Apple is reportedly partnering with Google for AI capabilities.

Key Arguments: OpenAI's pre-training runs have not been successful recently, leading to a need to catch up with competitors. Google's Gemini 3 is faster and more reliable, posing a significant threat to OpenAI's market share. Anthropic's Claude Opus 4.5 excels in style transfer and empathy, making it a strong enterprise tool. AI models are becoming commoditized, and distribution (e.g., Google's ecosystem) is increasingly important. AI slop can be both harmful (e.g., fake recipes hurting human creators) and benign or satirical (e.g., Bird Game 3). The 'California view' (what AI can do) vs. 'New York view' (what it can't) frames the debate on AI's utility.

Data Points: Gemini monthly users: 650 million - Google reported this number for Gemini usage. ChatGPT weekly users: 800 million - OpenAI reports weekly user numbers for ChatGPT. Anthropic annualized revenue: $9 billion - Anthropic expects this revenue by end of year, up from less than $1 billion. OpenAI projected revenue: $20 billion - OpenAI's projected revenue for the year. Apple's payment to Google for AI: $1 billion per year - Apple reportedly pays Google this amount for Gemini integration.

Pivotal Quotes: "I think the race is still very much on. The models are still very much getting better. It remains to be seen how soon or quickly that will kind of diffuse into products that actually make life look very different for people like you and me." — Kevin Roose: Discussing the ongoing AI competition and its impact on everyday life. "I've come to believe that there are fundamentally two different views of AI. There is what I call the California view of AI, which is what can it do? And then there's what I call the New York view of AI, which is what can't it do?" — Casey Newton: Framing the debate on AI's utility and criticism. "Slop in the name of love." — Casey Newton: Parting message in the 'Hard Fork Review of Slop' segment, advising to create slop that is good or neutral.

Implications: The AI landscape is highly competitive, with OpenAI facing pressure from Google and Anthropic. Users should experiment with multiple models as they rapidly improve. AI slop will continue to proliferate, requiring better detection and ethical considerations. Enterprise AI adoption is accelerating, particularly for coding and agentic tasks.