This piece really made me think: while the 'context is king' argument is incredibly compelling for current LLMs, I wonder if a completely new paradigm could emerge that shifts this fundamental dependency, and honestly, you've truely articulated this race so insightfully.
hey Jack, Really enjoyed this framing. Especially the idea of treating the context window like RAM and organising the whole system around preserving it for judgment.
I’ve been thinking along similar lines and recently wrote about context engineering through an “intent” lens.. not as memory management, but as a way to make agent behaviour debuggable and trustworthy.
curious to know if that framing lines up with what you’re seeing as agents get more capable.
This piece really made me think: while the 'context is king' argument is incredibly compelling for current LLMs, I wonder if a completely new paradigm could emerge that shifts this fundamental dependency, and honestly, you've truely articulated this race so insightfully.
hey Jack, Really enjoyed this framing. Especially the idea of treating the context window like RAM and organising the whole system around preserving it for judgment.
I’ve been thinking along similar lines and recently wrote about context engineering through an “intent” lens.. not as memory management, but as a way to make agent behaviour debuggable and trustworthy.
curious to know if that framing lines up with what you’re seeing as agents get more capable.