Loops explained: Claude, GPT, Mira and what works
The tweet discusses explanations of loops, comparing AI models like Claude and GPT, and evaluating what approaches are effective.
Background
This tweet discusses "loops" — AI research groups or companies (not software loops). Claude refers to Anthropic's AI assistant series (Claude 3, Claude 4). GPT is OpenAI's line of large language models (GPT-4, GPT-4o). Mira refers to Mira Murati, OpenAI's former CTO who led development of GPT-4 and DALL-E 3, and now heads her own AI venture. The post compares the philosophical approaches of these AI labs: Anthropic (Claude) focuses on safety and constitutional AI; OpenAI (GPT/Mira-era) emphasizes capability scaling; the tweet suggests evaluating what actually works in practice versus what's hyped. "Loops" here likely means teams/camps in the AI industry (a play on "loop" as in a group of people).