To manage context costs, Tasklet summarizes agent history with decreasing granularity over time. Recent interactions are sent verbatim, while older conversations have tool calls, thinking steps, and messages truncated or summarized. This is done in cache-aware buckets to minimize cost.

While prompt engineering is the interface, context engineering is the "magic" for production systems. It involves strategically managing what information (session history, knowledge base) fits into the model's limited context window. This art directly impacts both cost and performance.

The next frontier for visual intelligence is twofold: creating truly multimodal models that retain long-term context of user interactions without re-prompting, and developing real-time generation. Real-time capabilities are crucial for creating duplex interactions and enabling robots to perceive and act instantly.

Even models with million-token context windows suffer from "context rot" when overloaded with information. Performance degrades as the model struggles to find the signal in the noise. Effective context engineering requires precision, packing the window with only the exact data needed.

Astra achieves long-term task persistence not by summarizing its context window, but through a novel mechanism. It maintains a 'long lived notes file' that it can update and has the ability to search its entire session history, effectively managing a much larger context than its token limit implies.

Long-running AI agent conversations degrade in quality as the context window fills. The best engineers combat this with "intentional compaction": they direct the agent to summarize its progress into a clean markdown file, then start a fresh session using that summary as the new, clean input. This is like rebooting the agent's short-term memory.

Instead of just expanding context windows, the next architectural shift is toward models that learn to manage their own context. Inspired by Recursive Language Models (RLMs), these agents will actively retrieve, transform, and store information in a persistent state, enabling more effective long-horizon reasoning.

Seemingly complex features like long-term memory and skill creation are fundamentally clever systems for managing an AI's limited context window. The "harness" efficiently loads and unloads relevant information (memories, skills) at the precise moment it's needed, rather than keeping it all in context constantly.

To enable long-horizon tasks, Cursor incorporates "self-summarization" directly into its RL loop. The model learns to compact its own history and restart its context window with the summary. This allows it to operate over millions of tokens despite a nominal 200k context limit.