Small, focused experiments for learning how generative AI behaves.
Explore how a language model predicts likely next tokens and how temperature changes its suggestions.
Compare an LLM's arithmetic answer with JavaScript's exact result.
See how generative AI excels at producing content in formulaic contexts and what that reveals about its strengths and limits.
Ask the same question twice to see how a rude or grateful tone can shape an LLM response.
Route one question through two languages, then compare the answers after they round-trip back to English.
Explore how a model only sees messages that fit inside an adjustable estimated token budget.
Train through feedback in a local RLHF-style confidence simulation with no model training or network requests.
Challenge an AI assistant instructed to refuse anything related to milk or dairy.
Compare how Nova, Gemma, and DeepSeek reason through the same ethical dilemma.