Goal-binding default prompt + temp 0.2 (9/9 traps now catch); think-split in Trap-It-Yourself; accurate judge keywords; honest context claim

UI/UX redesign: tap-to-play trap card grid, animated verdict banner, stat-tile scoreboard, fixed hero/tab contrast

Submission metadata: track:wood, sponsor:nvidia, off-brand/off-grid/sharing badges

Parse <think> reasoning: show final answer, collapse reasoning, judge on answer

Rebuild as Gotcha Arena: reasoning-trap deck + scoreboard + Blind-Spot Check

Restore easy benchmark examples

Remove easy examples

Add goal-binding examples and 1k token default

Serve GGUF with persistent llama.cpp server

Run GGUF through official llama.cpp CLI

Build latest llama.cpp binding from source

Upgrade llama.cpp binding for Nemotron GGUF

Match Spaces runtime dependency

Enable real llama.cpp GGUF inference

Keep benchmark examples visible

Add benchmark-style prompt examples

Polish First-Principle AI interface

Point app to org model repo

Add First-Principle AI Gradio app

initial commit