Ju Lin's AI Weblog: ContextAn independent research notebook on AI engineering, agents, models and the systems around them.https://julin.ai/atom/tags/context/index.xmlJu Lin2026-10-23T00:00:00+13:00More Knowledge ≠ Smarter LLMhttps://julin.ai/2026/10/23/context-window-limitations/2026-10-23T00:00:00+13:002026-10-23T00:00:00+13:00<p>Putting more knowledge into an LLM’s context window does not make the model smarter. The context window is limited. Increasing knowledge in context degrades performance.</p>
<p>Only the relevant information matters—and only what’s needed for each turn.</p>
<p>I tested <a href="https://github.com/google-research/harness2">harness2</a> recently. Some tasks went from ~3 minutes to 30 minutes. Steps failed. The LLM needed more tokens to diagnose and retry. Overall success rate stayed the same.</p>
<p>The lesson: agent self-evolving is promising but has a long way to go. Carefully selecting what the agent sees is still more important than giving it everything.</p>