I’ve been a SuperGrok subscriber since before Grok Bots were introduced. I value the service, appreciate the innovation behind it, and remain optimistic about its potential.
That’s precisely why I feel compelled to speak up about the removal of editable instructions for personal Grok Bots.
I’ve invested considerable time developing a team of three specialized AI agents: a general-purpose coordinator, an independent fact-checker, and a creative collaborator.
These bots have different responsibilities, rules, and expectations for working with one another. I’ve spent time evaluating their performance, identifying weaknesses, and carefully refining their instructions.
This wasn’t just about customizing their personalities. It was about establishing a system where increasingly capable AI agents remain accountable to the person operating them.
Removing the ability to directly inspect and edit those instructions has effectively halted my ongoing development work.
I understand the appeal of simplifying the user experience and allowing people to configure their bots conversationally. That can be a valuable additional capability.
But why should it replace direct, transparent control?
An AI agent remembering something I told it is not equivalent to my being able to inspect, edit, and maintain the instructions that govern its behavior.
If the agent misunderstands an instruction, forgets something important, or begins behaving inconsistently, how am I supposed to reliably identify and correct the problem when I cannot directly examine its configuration?
This becomes especially concerning when multiple agents are expected to coordinate, delegate responsibilities, respect boundaries, and obtain human approval before taking consequential actions.
I have also read Cursor’s acknowledgment that instructions edited through the mobile interface may not be reliably applied to bot behavior. That raises concerns extending beyond the missing editor itself.
I’m not opposed to innovation or changes in how AI agents are managed. Quite the opposite. I’m excited about the possibilities, and I want these systems to succeed.
But I strongly believe that greater AI autonomy should be accompanied by greater transparency and meaningful user control, not less.
I’m asking Cursor and xAI to restore the ability to directly view and edit persistent instructions for existing personal Grok Bots, preserve established configurations, and provide clear information about how those instructions are applied.
Conversational configuration and direct editing should be complementary options, not mutually exclusive ones.
I have no intention of abandoning a service I otherwise value simply because I disagree with a product decision. But continued support for a product should not be mistaken for approval of every change made to it.
I hope the developers will reconsider this decision and engage with the users who have invested time building sophisticated systems around functionality that was previously available.
Please restore direct, editable bot instructions. Human oversight should be a fundamental capability, not an optional convenience.