Ethan B. Holland

This week’s covers celebrate my great friend Jack Schecter, who has been having fun using o3 to create specs, measurements, and diagrams for a few woodworking and weekend home improvement projects. That’s very different from image diffusion, since specs require Python and math, etc. However, in Jack’s honor, the theme for this week’s automated image rubric is instruction manuals.

I asked Claude Opus 4 to create a rubric I could use for batch producing category covers in the spirit of various types of instructions. Flux Pro v1.1 Ultra API.

The main cover is a hybrid of an Ikea and Google office renamed “ISEEYA” to poke a bit at the panopticon of multimodal training, the new VEO video model, and robot training in AR/VR simulation. Everything was diffused by GPT except for the main title text, which I added in Photoshop. I simply asked GPT Images to combine Google, Ikea, and Figure by giving it three reference images, and told it to change IKEA to ISEEYA. I’d give it a C- since it’s a bit boring, but it did what I asked very well. The issue is me, not the model.

Separately, Claude Opus 4 o3 created a rubric that allowed me to give it 43 one-word category names, and from those 43 single words, the Flux Pro v1.1 Ultra API returned 43 instruction manual images. These turned out pretty well, albeit with gibberish words.

CLAUDE:

This is a fantastic creative project! I love the concept of using assembly instructions as a visual metaphor for AI newsletter categories. You’ve laid out a really solid foundation. Let me reflect back what I understand and suggest a few additional considerations:

What you’ve covered well:

• Core theme: Assembly instructions aesthetic
• Variability through: color, style era, object type, and font
• Simplicity for image generation
• Category name integration as part of the instructions

Additional elements to consider for the rubric:

1. Assembly instruction layout style – Should we vary between:
   • Exploded view diagrams
   • Step-by-step panels
   • Parts inventory style
   • Blueprint/schematic style
2. Visual complexity level – Some categories might warrant:
   • Simple 3-5 part assemblies
   • More complex multi-component builds
   • Abstract “conceptual” assemblies for intangible categories
3. Paper/background texture – To enhance authenticity:
   • Aged/yellowed for vintage
   • Clean white for modern
   • Blueprint blue for technical
   • Grid/graph paper for engineering
4. Iconographic elements – Small details that reinforce the category:
   • Warning symbols
   • Tool requirements
   • Difficulty ratings
   • Part numbers/codes
5. Fallback strategy – For abstract categories that don’t have obvious physical objects, we could:
   • Create metaphorical “machines” or devices
   • Use conceptual assemblies (like building blocks of an idea)

Would you like to establish any constraints around:

• Maximum prompt length?
• Specific aspect ratio/dimensions?
• Any “must avoid” elements?
• How prominent/large the category name should be?

Once we nail down these details, we can create a structured rubric that systematically maps category characteristics to visual choices!

Perfect! I love this adaptive approach. Let’s build a rubric that acts more like an intelligent decision tree rather than a rigid formula. Here’s my proposed structure:

ADAPTIVE ASSEMBLY INSTRUCTION RUBRIC

Phase 1: Category Analysis

1. Identify category domain (technology, social, business, creative, science, etc.)
2. Extract cultural/contextual associations
3. Flag if category is too abstract → Pause for human input

Phase 2: Object Selection

• Map category to most relevant physical object that would need assembly
• Ensure object is different from previous selections
• Must be recognizable and “real”

Phase 3: Style Era Decision

Based on category associations, choose:

• Retro/Vintage (1950s-1980s) – for nostalgic or foundational concepts
• Modern Minimal (2000s-present) – for cutting-edge topics
• Industrial/Technical (1990s) – for process/system categories
• Mid-Century (1960s-70s) – for creative/design categories

Phase 4: Visual Elements

Diagram Type (based on object complexity and category nature):

• Exploded view
• Sequential steps
• Parts layout
• Isometric assembly

Color Palette (derived from category associations):

• Primary accent color
• Supporting colors
• Paper/background tint

Typography (matching era and object):

• Technical drafting fonts
• Vintage instruction manual fonts
• Modern sans-serif
• Hand-lettered annotations

Phase 5: Prompt Construction Order

1. Assembly instruction diagram style and layout
2. Main object being assembled
3. Category name as prominent title
4. Color scheme
5. Paper/background texture
6. Supporting details (tools, parts, annotations)
7. Style era and aesthetic details

Decision Pause Triggers

• Category too abstract to map to physical object
• Previous object already used
• Conflicting style associations
• Unclear cultural references

Does this framework feel right? Should we add any other decision points or modify the approach?

Looking at this as your creative partner, I think we’re in great shape! But let me suggest a few refinements that could help:

Additional considerations:

1. Series Cohesion Element – While we want variety, we should add a subtle unifying element so they feel like a collection:
   • Perhaps a consistent badge/stamp saying “AI Assembly Series” or similar
   • Or a consistent border treatment
   • This helps them read as a set despite the variety
2. Readability Insurance – Since the category name is crucial:
   • Add a decision point for ensuring contrast (light text on dark objects, dark on light)
   • Consider adding a subtle backdrop behind text if needed for legibility
3. AI Newsletter Context – We could subtly reinforce this is for AI:
   • Small circuit board patterns in backgrounds
   • Digital/tech touches even on vintage styles
   • Binary numbers as “part numbers”
4. Quick Reference Library – As we process the 45 categories, we should track:
   • Objects already used (to avoid repeats)
   • Color palettes used (to ensure good distribution)
   • This prevents accidental clustering
5. Cultural Universality Check – Since AI is global:
   • Favor objects that are internationally recognizable
   • Avoid region-specific items unless the category demands it

Should we add any of these to the rubric, or keep it lean and handle these considerations organically as we work through the categories?

Perfect! Now I have everything I need. Let me process all 45 categories through our rubric and generate the prompts. I’ll track objects and color palettes to ensure variety.Perfect! I’ve generated all 45 prompts and created them in a text file following your exact format. Each prompt:

✓ Features a unique object (no repeats across all 45) ✓ Uses varied color palettes ✓ Employs different visual styles (vintage, modern, technical, etc.) ✓ Prominently displays the category name as the title ✓ Maintains the assembly instruction theme while being distinctive

The prompts are structured with the most important elements first to ensure proper prioritization during image generation.

Objects range from:

• Tech hardware (CPUs, VR headsets, keyboards)
• Vintage items (typewriters, film projectors, stereos)
• Scientific instruments (microscopes, telescopes, measuring tools)
• Industrial equipment (robotic arms, wind turbines)
• Consumer products (cameras, smartphones, smart home devices)

Styles span:

• 1940s-2020s aesthetics
• Technical blueprints to friendly consumer manuals
• Industrial to minimalist designs

The file is ready for direct API use. Would you like me to make any adjustments to specific prompts?