OpenAI just dropped a research preview of GPT-5.3-Codex-Spark, a smaller, faster version of their code model built for real-time collaboration. It’s the first fruit of their partnership with Cerebras, delivering over 1000 tokens per second on ultra-low latency hardware. Think of it as the code companion that keeps up with your pace—you can interrupt it mid-thought, redirect it, or ask for tweaks and see results instantly. While it’s text-only with a 128k context window for now, it’s a taste of a bigger vision: combining long-running autonomous coding sessions with instant, interactive editing. ChatGPT Pro users get first dibs to experiment, though you might hit rate limits or queues as they balance demand. Why it matters: Codex-Spark signals a shift toward AI that works with you, not just for you—where response time is as crucial as raw intelligence, making real-time pair programming with AI feel like a natural extension of the developer’s workflow.