Skip to content
View maximbex's full-sized avatar
7 followers · 43 following

Block or report maximbex

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
maximbex/README.md

Hi, I’m Max

I’m interested in projects that help structure information technology either based on silicon substrate or biological substrate. My background touches on AI, epistemology, systems design, the philosophy of science, theology, and law — and I'm curious about how these topics can bidirectionally inspire better infrastructure.

  • Currently exploring: Agentic Systems and the preconditions for effective recursive self-improvement, semantic knowledge systems, and research-oriented tooling.
  • Familiar with: Java, HTML/CSS/JS, LaTeX, Python, FileMaker Pro, and MacOS/Linux/Windows/Synology environments.
  • Open to collaborate on: Any project that focuses on meta-strategic implementation of ethically consistent value generation.

Reach me: GitHub Issues / Discussions for now. Fun fact: I have a forklift driver's license which I used last time, during a student job (sometimes the most satisfying way to bridge abstraction and implementation is just moving heavy things around in the physical world).

Favorite Quote (of the day): "Premature optimization is the root of all evil" (Donald E. Knuth)

Pinned Loading

  1. event_horizon_integrity_gist_draft.md event_horizon_integrity_gist_draft.md
    1 
    # Event-Horizon Integrity: Reconstructible Metadata for Preventing Interactional Hallucination in LLM Training
    2 
    
              
    
              
    
                3
                
    ## Abstract
    
              
    
              
    
                4
                
    
              
    
              
    
                5
                
    Metadata is widely used in language-model training to specify task difficulty, risk, expected reasoning styles, and safety boundaries. However, when metadata contains claims that cannot be reconstructed from runtime-accessible sources, it introduces structural train-runtime mismatch. During training, the model learns to rely on invisible interpretive labels (oracles). At runtime, these labels are absent, forcing the model to hallucinate the interpretive frame of the exchange — a failure mode we call **interactional hallucination**. 
    
              
    
          
    
    
    
  
    

    2. 

      

  2. 
  
    
    
    
      
    
        
        A daily reference and 9-step recolle...  A daily reference and 9-step recollection practice for calibrating attention, aligning systems with their true purpose, and resisting metric corruption.      
    

          
    
              
    
                1
                
    # Superiority Without Self-Exaltation
    
              
    
              
    
                2
                
    
              
    
              
    
                3
                
    *A daily reference for recollection, discernment, and metric purification*
    
              
    
              
    
                4
                
    
              
    
              
    
                5
                
    ## 0. Core Orientation
    
              
    
          
    
    
    
  
    

    3.