Replit CEO Masad宣布OpenAI取代Anthropic成为平台默认模型，Claude仍可选。
Replit officially designated OpenAI as its default large language model provider, replacing Anthropic’s Claude series. This shift reflects real-world usage patterns—not a full deprecation, but a prioritization based on latency, reliability, tooling integration, and developer feedback. Claude remains accessible for users who explicitly opt in. The move signals a broader trend: AI application platforms are converging on proven, production-ready APIs rather than maintaining parity across multiple models for theoretical flexibility. For early-stage AI startups, this underscores a critical infrastructure lesson: deep coupling to any single proprietary model’s behavior (e.g., specific system prompt handling, function-calling syntax, or context window quirks) creates significant technical debt. Teams should architect model abstraction layers from day one and validate fallback paths—even during MVP—especially when targeting enterprise customers, for whom model portability is often a non-negotiable procurement requirement.