Added documentation of Didactopus helping reduce 'offloading'

README.md

@@ -4,6 +4,8 @@
 
 Didactopus is a local-first Python codebase for turning educational source material into structured learning domains, evaluating learner progress against those domains, and exporting review, mastery, and skill artifacts.
 
+Its intended use is closer to a structured mentor or self-study workbench than a "do my assignment for me" engine. The project should help learners get guidance, sequencing, feedback, and explanation without encouraging the offloading effect that comes from unstructured GenAI use.
+
 At a high level, the repository does five things:
 
 1. Ingest source material such as Markdown, text, HTML, PDF-ish text, DOCX-ish text, and PPTX-ish text into normalized course/topic structures.
@@ -36,6 +38,8 @@ That gives you:
 - progress artifacts
 - a reusable skill grounded in the exported knowledge
 
+The point is not to replace your effort. The point is to give your effort structure, feedback, and momentum.
+
 If that is your use case, read the next section, `Fast Start For Impatient Autodidacts`, and skip the deeper architecture sections until you need them.
 
 ## Fast Start For Impatient Autodidacts
@@ -116,6 +120,17 @@ Treat Didactopus as a loop:
 
 The important idea is not "perfect ingestion first." It is "usable learning structure fast enough that you keep going."
 
+### If you are using it alongside coursework
+
+The intended pattern is:
+
+1. Use Didactopus to clarify the topic map and prerequisites.
+2. Ask it for hints, sequencing, comparisons, and self-check prompts.
+3. Use its outputs to diagnose where you are weak.
+4. Still do the actual writing, solving, and explaining yourself.
+
+That is the difference between assisted learning and offloading. Didactopus should help you think better, not quietly substitute for your thinking.
+
 ### Current friction honestly stated
 
 The lowest-friction path is the included demo. The custom path still asks you to be comfortable with:
@@ -139,6 +154,8 @@ Instead of only reading notes, you can get:
 - evidence-aware progress artifacts
 - reusable skill outputs for future tutoring or evaluation
 
+In the best case, that makes learning feel more like active skill-building and less like either passive consumption or answer outsourcing.
+
 ## What Is In This Repository
 
 - `src/didactopus/`

docs/faq.md

@@ -4,6 +4,26 @@
 
 Didactopus turns educational material into structured learning packs, then uses graphs, evidence, and review workflows to support human or AI learning against those packs.
 
+## Is this meant to help me learn, or to do the work for me?
+
+It is meant to help you learn.
+
+The intended role is:
+
+- clarify topic structure
+- surface prerequisites
+- suggest study order
+- provide explanations, comparisons, and self-checks
+- help you see where your understanding is weak
+
+The intended role is not:
+
+- silently complete coursework for you
+- replace the need to explain ideas in your own words
+- turn learning into answer copying
+
+In other words, Didactopus is supposed to reduce confusion and friction without encouraging the offloading effect of unstructured GenAI use.
+
 ## Is this a packaged application or a research/workbench repository?
 
 It is a workbench-style repository with runnable code, tests, example packs, generated outputs, and local-first review/demo flows.
@@ -93,6 +113,18 @@ Right now the value is in:
 
 The current demos show the shape of a mentor workflow even though the agent itself is not yet a live external model integration.
 
+## How should I use it if I am taking a course and do not want to hire a tutor?
+
+Use it as a structured study companion:
+
+1. Build or load a topic pack.
+2. Use the path and prerequisite structure to see what to study next.
+3. Ask for hints, comparisons, and explanation prompts.
+4. Use progress artifacts to identify gaps.
+5. Do the actual solving and writing yourself.
+
+That keeps the system on the "guided practice" side of the line instead of the "outsourced thinking" side.
+
 ## What is the current evidence model?
 
 The evidence engine supports: