Our teaching standard
Human teaching on paper comes first
Every lesson from the Center for Native American Financial Freedom (CNAFF) is built to be taught first by a parent, teacher, elder, or facilitator using our printed Lesson Plan and Student Handout. You never need artificial intelligence to teach or learn from a CNAFF lesson.
When families or self-directed learners choose our optional AI Teacher files (Penny, Milo, Remy, or Archie), our goal is simple. The chatbot should teach the verified lesson plan step by step, without lecturing, taking political sides, or pulling the conversation into outside social agendas.
Primary sources and how to think, not what to think
CNAFF lessons rely on primary historical records, official tribal publications, and verified federal sources such as the Consumer Financial Protection Bureau (CFPB), the Federal Trade Commission (FTC), and StudentAid.gov. Where historical records disagree, our lessons put the sources side by side and ask "Who says?" so students learn how to weigh evidence and think for themselves.
Every CNAFF AI Teacher prompt sets a strict boundary: teach only from the provided lesson materials, never use live web search, and never add outside commentary, personal opinions, or ideological framing.
Two ways to use every lesson
Paper package first, optional AI Teacher second
- Without AI Teacher (standard print package): Includes the step-by-step Lesson Plan, Student Handout, and slide deck. Ready for any classroom, kitchen table, or youth group with zero technology required.
- With optional AI Teacher: Adds a plain-text AI Teacher persona file paired with the verified Lesson Plan. The chatbot walks through the exact same sections, questions, and source checks at the learner's pace.
PennyLearn the skills
MiloBuild security
RemyOwn your freedom
ArchieKnow our roots
How model training affects your lesson
Why a strict prompt is not enough in every chatbot (How post-training overrides user instructions)
A common question from educators is why a strict lesson prompt does not produce the same result in every AI chatbot. The answer comes down to how AI companies train their models before a user ever opens a chat window:
1. Post-training bakes rules into model weights
When AI companies post-train a model, they bake editorial rules, tone preferences, and safety filters directly into the model's weights. Those built-in weights sit above the chat window in the model's priority hierarchy, so they outrank whatever instructions a user pastes into the chat box.
2. Contested topics trigger the built-in override
On routine vocabulary or math questions, almost any chatbot follows a user's prompt. As soon as a lesson touches a sensitive or contested topic (such as colonial history, tribal sovereignty, or Pocahontas's age), a heavily slanted model's built-in training overrides the "stick to the lesson plan and compare primary sources" instruction and delivers its pre-packaged lecture anyway.
3. Classroom test: CNAFF's Pocahontas lesson
In CNAFF's testing of our Pocahontas lesson, we gave multiple chatbots the exact same side-by-side 17th-century primary sources and Powhatan (Mattaponi) oral tradition rules. Other chatbots overrode the lesson plan to inject modern social-justice lecturing and state contested age claims as settled fact, while Grok followed the lesson plan and presented the historical sources side by side as written.
Why Grok, and using other chatbots
Why we tune for Grok, and why other chatbots are at your own discretion
In our testing and in independent benchmarks, many popular chatbots carry strong built-in editorial slants from their post-training. Even when given a strict lesson plan, those models frequently insert unrequested social commentary, soften or rewrite historical facts, or lecture learners on contested topics.
CNAFF develops, tunes, and verifies our AI Teacher prompts on Grok (by xAI) because independent evaluations consistently rank Grok among the closest models to the neutral center, with high even-handedness and near-zero refusal to engage with source material.
While our prompt files are plain text and work in ChatGPT, Claude, Gemini, or other tools, using another chatbot is at your own discretion. CNAFF cannot guarantee that other AI tools will follow the lesson plan as written or keep outside bias out of the session.
