Android Authority Writer Says Claude’s Logo Design Failed at Print Scale

The writer said he has paid for the $20 Claude Pro plan for months and uses it for personal vibe-coding projects, including a custom screenshot-framing tool for articles and a Claude Code automation that pulls interesting links and pitches from his inbox. Those tasks worked, so he tried a more ambitious project while setting up his wife’s dental practice.

The practice’s branding and logo are important, and the couple had a designer, but she was unavailable for a week or two while they needed to decide on outdoor signage. In India, the writer said, that process involves board designers, technicians and printers with their own requirements and limitations. He already had a wordmark and drawings for the logo and held a decade-old UX design diploma, so he decided to use Claude.

He uploaded the wordmark in SVG format, provided a rough sketch of a tooth-based logo incorporating a “T” motif and wrote a detailed design prompt. The first attempt was far off. After instructions to thin out parts and change curves, and after using 20% of his five-hour usage limit over about 10 minutes, he had something decent. Claude gave four ideas; they picked one and continued. After about an hour, even his wife was impressed.

The writer was using Claude around the release of Opus 5 and Fable 5 and had free Fable 5 credits, which he quickly depleted in repeated edits. He and his wife kept asking for adjustments: longer roots, a shorter valley between the teeth, changes to the crown’s curvature and more. At one point, Claude made a slider-based tool that let them control the distance of the root valley and the length of the “T” motif. They dialed it in and asked Claude for a mockup of the signage using their wordmark. Satisfied with the colors and look, they sent it to their designer and the signage person.

Problems appeared as soon as the mockup was sent. The font size would not work and needed to be bigger and thicker, but Claude could not generate a thicker font that looked the same as the original, the writer said. The generated wordmark lost the font’s sharp characteristics and rounded the edges, even after more token use. Claude also gave instructions about dealing with signage technicians and design options such as CNC cutouts versus 3D embossed lettering, but the technical dimensions it provided were far off. The account said the AI presented the information confidently.

The issues became clearer when the designer opened the file. The logo looked fine as a small icon, but zoomed in it was filled with jagged edges. After two days of prompting, the writer said, Claude still could not smooth the curve at the crown of the tooth. It was passable for an Instagram logo but impossible to print at four feet tall. The more he prompted, the longer the context window became and the worse the results got.

The writer described serving as a middleman between a technician and an AI model and tried to redo the entire project. He asked Claude to generate the whole thing on its own, hoping changes would then be easier, but called that a horrible idea; the outputs would not have passed a first-year college class. His designer returned to town, and they rushed to sit with her. In about an hour, she spoke with the technician, understood the details that needed changing and made the font thicker while keeping the edgy look, manually adjusting the design.