Hi HN,

  I built RLM Analyzer, an open-source tool for analyzing codebases using       
  Recursive Language Models.                                                    
                                                                                
  Problem: LLMs have context limits. Large codebases don't fit.                 
                                                                                
  Solution: Based on MIT CSAIL's RLM paper (arXiv:2512.24601), the AI writes    
  code to explore files and delegates analysis to sub-LLMs. Key innovations:    
                                                                                
  - Context compression (50-70% token savings)                                  
  - Sliding window history                                                      
  - Memory bank for key findings                                                
  - Adaptive compression based on usage                                         
                                                                                
  Example commands:                                                             
    rlm summary          # Codebase overview                                    
    rlm security         # Vulnerability scan                                   
    rlm arch             # Architecture analysis                                
    rlm ask "How does auth work?"                                               
                                                                                
  Also works as an MCP server with Claude Code/Cursor, saving 60-73% on API     
  costs by offloading to Gemini.                                                
                                                                                
  GitHub: https://github.com/zendizmo/rlm-analyzer                              
  NPM: https://www.npmjs.com/package/rlm-analyzer                               
                                                                                
  Built with TypeScript, uses Gemini API. Feedback welcome!