AJ Isaacs c63913cf14 fix: anonymize usernames before LLM analysis to prevent name-based scoring bias ...

Display names like "Calm your tits" were causing the LLM to inflate toxicity
scores on completely benign messages. Usernames are now replaced with User1,
User2, etc. before sending to the LLM, then mapped back to real names in the
results.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

2026-02-25 22:20:53 -05:00

..

analysis.txt

fix: anonymize usernames before LLM analysis to prevent name-based scoring bias

2026-02-25 22:20:53 -05:00

chat_chatty.txt

Add context format explanation to chat prompts

2026-02-23 09:54:08 -05:00

chat_drunk.txt

Add drunk mode: happy drunk commentating on everything

2026-02-23 20:05:03 -05:00

chat_english_teacher.txt

Tune english teacher to catch more errors, bump roast reply chance

2026-02-23 11:03:03 -05:00

chat_hype.txt

Add hype mode: positive/supportive teammate personality

2026-02-23 20:02:39 -05:00

chat_personality.txt

Add context format explanation to chat prompts

2026-02-23 09:54:08 -05:00

chat_roast.txt

Tone down roast bot: more positive, less frequent

2026-02-23 19:55:17 -05:00

scoreboard_roast.txt

Generalize image roast to handle selfies, memes, and any image

2026-02-23 12:15:22 -05:00