Course 1 · Tier 4
BotSpeak: The Nine Pillars of AI Fluency
How to talk to machines — and know when they're talking past you. This is the series entry point: the course that establishes AI collaboration as a professional competency, not a novelty.
The Five Modes Framework
BotSpeak organizes AI interaction into five distinct modes of engagement. Each mode demands a different kind of human judgment — from formulating precise queries to evaluating whether the output actually answers the question you meant to ask.
The framework treats prompt engineering not as a trick or a hack, but as a communication discipline. Students learn to recognize when a model is confidently wrong, when it's hedging because it doesn't know, and when the output is structurally sound but semantically empty.
- Mode 1: Directed Query — asking the right question in the right form
- Mode 2: Iterative Refinement — shaping output through structured dialogue
- Mode 3: Constraint Specification — defining what the model must not do
- Mode 4: Evaluation and Audit — assessing output against human standards
- Mode 5: Collaborative Synthesis — combining human and machine contributions
The Tier Taxonomy
Not all human capacities are equally resistant to automation. The tier taxonomy provides a framework for assessing which skills AI can approximate, which it can assist, and which remain entirely beyond its reach.
BotSpeak introduces this taxonomy at Tier 4 depth — meaning students engage with the material at a level that requires genuine critical analysis, not just recall or application. The taxonomy itself becomes a tool for evaluating every subsequent course in the series.
Hallucination Detection
Models produce plausible-sounding outputs that are factually wrong. Students learn systematic methods for identifying hallucinations — not by checking every fact, but by recognizing the structural patterns that indicate confabulation.
The Rhetorical Gap
The distance between what you intend to communicate and what the model interprets. BotSpeak trains students to close this gap — not by dumbing down their prompts, but by understanding how models parse intent.