Power Users Report Better Results After Disabling Memory in Claude and ChatGPT
A growing number of heavy users are finding that turning off persistent memory features in major AI assistants produces noticeably better output — raising questions about whether memory systems are creating more problems than they solve.
@sriramenn posted a concise observation that resonated with power users: "i've turned off all memory settings in claude desktop / chatGPT and my experience is notably better." The claim is straightforward, but its implications are significant. Memory features — where the model retains context about you across sessions — have been marketed as a key differentiator by both Anthropic and OpenAI. If heavy users are finding these features degrade performance, something is wrong with either the implementation or the premise.
Unlock the full briefing
Get every story in today's briefing, the full archive, and the daily AI intelligence brief.
All stories today
Full archive
Daily brief
Cancel anytime. Payments powered by Stripe.