Long Context Tests: Infinite Scroll Works Well in Personal Agents

vivekhaldar · x · 2026-08-29

Vivek Haldar questions whether concerns about 'lost in the middle' and context rot are still valid. Based on his personal experience using infinite scroll within agents, it seems to work fine, though he notes this is anecdotal. He suggests that a key reason for success is the main thread's ability to spawn sub-agents to handle heavy lifting.

Related event: Long Context Works Well in Practice as Sub-Agents Share the Load(2 posts)→

Original post →

More from coding & agent

coding & agent channel →