Why can’t LLMs just LEARN the context window?

Check out HubSpot's FREE 2026 Guide to AI Agents: https://clickhubspot.com/3972be In this video, I'll be breaking down a new approach to long-context LLMs called test-time training (TTT-E2E), where models store past context directly in their weights instead of relying on attention or KV caches. Kind of like meta learning, but with gradient descent. my latest project: Intuitive AI Academy We just wrote a new piece on MoE! https://intuitiveai.academy/ limited time code "EARLY" for 40% off yearly plan! TTT-E2E [Paper] https://arxiv.org/abs/2512.23675 Appeared papers [Titans] https://arxiv.or…