Keyboard shortcuts

Press ← or → to navigate between chapters

Press S or / to search in the book

Press ? to show this help

Press Esc to hide this help

Lesson 4: AI-Enhanced Verification and Multimodal Feedback

Estimated time: 45-60 minutes

Learning Objectives

  • Run 3dm info to collect deterministic renders and AI diagnostics (if configured)1 2
  • Compare AI suggestions with renderer and slicer outputs and prioritize deterministic fixes3
  • Record and sanitize AI prompts and outputs for reproducibility and data governance3

Materials

  • 3dMake configured with an LLM API key (optional)
  • Example project with renderable geometry

Step-by-step Tasks

  1. Verify API configuration (if used) or run 3dm info --dry-run to confirm render pipeline works locally1.
  2. Run 3dm info and save the produced images and textual report to build/1.
  3. Inspect deterministic outputs (render warnings, slicer preview) and compare them to AI recommendations; prioritize deterministic issues3 2.
  4. Iterate prompt engineering (in 3dmake.toml or via --prompt) with precise technical primitives and re-run 3dm info to examine changes3.
  5. Document all prompts, AI outputs, and deterministic validation steps in AI-notes.md within the project3.

Checkpoints

  • After step 2 you have stored render images and the AI report in build/1.

Quiz - Lesson 3dMake.4 (10 questions)

  1. What command generates AI diagnostics and model renders1?
  2. Why must AI outputs be validated against renderer/slicer results3 2?
  3. Name one privacy or governance concern when sending models/images to an API3.
  4. What is an example of a technical primitive to include in a prompt3?
  5. Where should you record prompts and AI outputs in the project3?
  6. True or False: AI-generated suggestions should always be implemented without verification.
  7. Describe what “deterministic validation” means in the context of AI-assisted 3D design.
  8. Explain how you would use 3dm info to validate AI suggestions against your actual model.
  9. What steps would you take if the AI suggests a design change that conflicts with your requirements?
  10. How would you document AI-assisted decisions in your project for reproducibility and transparency?

Extension Problems (10)

  1. Create an AI-notes.md documenting three prompts and the AI’s responses; indicate which suggestions you acted on3.
  2. Simulate a false-positive AI warning: describe how you validated and rejected it using deterministic checks2.
  3. Generate a short prompt that requests the top three structural risks and record the results3.
  4. Create a short checklist for sanitizing uploads before sending to an API3.
  5. Re-run 3dm info after a code fix and compare the differences in the AI report1.
  6. Design a verification workflow: use AI suggestions + manual validation + physical testing to validate design decisions.
  7. Create a transparency log for AI-assisted features: document all AI interactions, suggestions adopted, and decisions made.
  8. Build an AI-assisted design case study: from initial prompt through final print; document the entire decision process.
  9. Develop guidelines for trustworthy AI use in 3D printing: when to trust AI, when to verify, and how to validate recommendations.
  10. Write an accessibility guide for using AI features in 3DMake: how to interpret AI suggestions non-visually and understand confidence levels.

  1. 3DMake GitHub - https://github.com/tdeck/3dmake ↩ ↩2 ↩3 ↩4 ↩5 ↩6

  2. Deterministic Validation in Design - https://www.nist.gov/publications ↩ ↩2 ↩3 ↩4

  3. AI Output Validation - Best Practices for Prompt Engineering - https://en.wikibooks.org/wiki/OpenSCAD_User_Manual/The_OpenSCAD_Language ↩ ↩2 ↩3 ↩4 ↩5 ↩6 ↩7 ↩8 ↩9 ↩10 ↩11 ↩12