cpeterso

What tools or AGENTS.md prompts do you find effective for humanizing LLM output? Default GPT output is acceptable, but Claude is unbearable.

The “Simplified Technical English” fad was pretty effective, but now I just specify:

  - Use short sentences and active voice.
  - NEVER use the phrases "load-bearing", "seam", "unlock", "real", or “learnings”.

Maybe new models training on the previous generation’s synthetic output is compounding the effect? LLM Ouroboros.

I find Claude’s writing unbearable. That ChatGPT offers personalization options (Base style, warmth, enthusiasm) and Claude doesn’t suggests Anthropic doesn’t recognize this as a problem. Plus, people sharing ChatGPT writing won’t all have the same voice.

It would be cool to leverage that Street View driving capability to build a multiplayer Gumball Rally/Cannonball Run cross-country car race. Are there continuous routes recorded from New York City to Long Beach? Which is fastest? How long would that virtual drive take in Street View?

You’re right. I looked it up: getting the number of files in a directory in FAT and FAT32 file systems is O(n). I had assumed the count would be recorded in an inode-like structure but instead the file system must scan a directory table to count the non-null entries.

Raymond’s selectRandomFromIterator algorithm requires iterating over all the files. It seems like querying the file system for the number of files in a directory should be an O(1) operation. Then you just select random number N between [1, number of files] and iterate to the Nth file.

Why does Raymond assume counting the files is an O(n) operation?