Project: Maestro - Substack Newsletter Overall Performance
Using AI to design and develop a consumer application
CoPilot + Microsoft Designer = Winner
My overall experience generating both a newsletter logo and a title / wordmark was satisfactory. While AI accelerated the ideation phase, the tools struggled to produce simplified graphics.
Results
Ideation and Design
Ideation and theme design worked well. I noticed a high degree of AI group think in the suggestion of theme elements, colors and potential project names. Overall, AI tools would be useful for setting a baseline during the spitball phase of ideation, but you would still need human creativity to build a truly engaging concept.
Notes
- CoPilot (A+) - Tied with ChatGPT but gets an A+ for auto generating the winning image from chat
- ChatGPT (A) - Tied with CoPilot usable output
- Bard (B) - A+ for logo design but a C for wordmark, hit or miss
- Claude.Ai (C) - Regurgitated ideas, abbreviated original prompts, no real improvement on original request or themes.
- Grok (F) - Regurgitated my requests or provided basic responses. Not very creative even in creative mode.
Image Generation
If this were a professional project, I would outsource graphics design on a gig site like Fiverr. Under no circumstances would I put these AI generated images in front of a client or investor. However, for a project or prototype they are OK as placeholders.
Notes
- CoPilot + Microsoft Designer (A+) - Effectively used enhanced Co-Pilot output and generated a usable design. Designer enhanced and animated animated the image efficiently.
- Any + Dalle-E (C) - Struggled to produce a useable design, text was garbled and unusable (better to add text later, but seriously impacts ability to generate typography)
- Any + Stable Diffusion (F) - Fails on most abstract image requests
Note: Gap between models is in the literal interpretation of requests