The rise of AI coding assistants has created a new frontier in software development, with developers increasingly sharing their experiences with tools like Cursor AI. A recent discussion around .cursorrules—configuration files that guide AI behavior—reveals both the promise and frustration of this emerging technology.
The Art of AI Instruction
The community discussion highlights how developers are crafting increasingly detailed instructions for AI coding assistants, sometimes even resorting to aggressive language to get desired results. One developer's .cursorrules file begins with DO NOT GIVE ME HIGH LEVEL SHIT followed by specific instructions about code commenting practices. This approach has sparked debate about the current state of AI tools, with some users finding it bleakly funny that such forceful language is necessary to achieve reliable results.
These experiences reflect a growing recognition that effective AI collaboration requires explicit guidance. Developers are creating elaborate rule sets to prevent common AI behaviors like removing code comments, generating overly generic responses, or ignoring project-specific coding styles.
Evolving Prompt Strategies
More experienced users have developed sophisticated approaches to .cursorrules files. Rather than focusing on tone or output format, they use these files to provide project context, architecture details, and specific build instructions. One developer explained their strategy:
I use it to explain project-specific details so the model doesn't have to re-figure out what we're doing at the start of every conversation... It's critical to teach the model how to run the build+run feedback loop so it can fix its own mistakes without your involvement.
Some developers include verification mechanisms, such as requesting the AI refer to them as boss to confirm the rules have been properly processed. When the AI stops using this term, it signals that the context window may be overloaded or the conversation has grown too lengthy.
Common .cursorrules Strategies:
- Project context explanation
- Build and test instructions
- Language/library version specifications
- Verification mechanisms (e.g., "refer to me as 'boss'")
- Code comment preservation requirements
- Response format preferences
The Cursor vs. Copilot Debate
The discussion reveals an ongoing comparison between GitHub Copilot and Cursor, with developers weighing the merits of each tool. Cursor's YOLO mode—which allows the AI to autonomously write code, run tests, and make corrections—emerged as a significant differentiator. Supporters claim this feature enables developers to describe a feature, confirm a plan, and return minutes later to functioning code with passing tests.
However, skeptics question the quality of AI-generated tests, noting examples of trivial assertions that provide little value. This highlights a broader concern about whether AI tools genuinely increase productivity or merely create an illusion of progress that requires significant human intervention to refine.
Cursor vs. GitHub Copilot Features:
- Cursor: Chat window with "YOLO mode" for autonomous coding
- Cursor: Project-specific rules files
- Cursor: Ability to run build/test cycles independently
- Copilot: Integrated with GitHub ecosystem
- Copilot: Generally less expensive for GitHub Pro users
The Ritual of AI Interaction
Perhaps most tellingly, the community discussion reveals how interacting with AI coding assistants has begun to resemble ritualistic behavior. Users share unusual but effective prompts, such as telling the AI to go outside on the front porch and hit your weed vape, look at some clouds and trees, then come back inside and try again when it gets stuck in a loop.
This anthropomorphizing of AI tools has led to humorous comparisons to religious practices, with one user noting we are halfway to holding elaborate rituals to appease the LLM spirit. Another added that the prayers and vigorous side-smacking are already here; only incenses and sanctified oils remain to be seen.
As AI coding assistants continue to evolve, developers are adapting their workflows and expectations. While some embrace these tools as revolutionary aids that enhance productivity, others remain skeptical about the long-term impact on code quality and developer experience. What's clear is that effective collaboration with AI requires more than technical knowledge—it demands a new kind of communication skill that bridges human intent and machine understanding.
Reference: Update .cursrorrules about code comments