Claude Design rocks and Claude Design sucks. Let's talk about. Last night, I created an entire game in Claude Design inspired by the Longyou…

Claude Design rocks and Claude Design sucks. Let's talk about. Last night, I created an entire game in Claude Design inspired by the Longyou Caves in China (why is there no soot!), and we all gathered around the TV and Airplayed it and shouted at the screen to solve ciphers. Was it fun? YES! Was it perfect? God no! Claude is still rough around the edges. Here are some weak spots we found and what I would go to another product for, have to work extra hard to get done, or hire a human to create: 1) anything SVG from scratch (ex: designing caves or bat characters) - the illustrations Claude makes are comically bad. The "bat with a flashlight" we asked for looked more like a fat clown fish stuck in a purple hat wearing a misshaped pearl necklace. 2) accessibility decisions, like font sizes or color schemes - if you prompt Claude to fix these issues, it can go online and look up accessibility best practices and update the design decently well. But as a colorblind person who wears decently strong contacts, I would trust a human more than AI on this any day of the week. One example it consistently failed on was deciding what color to highlight text with. We could never read the text that was highlighted. 3) basic design principles - incredibly strange font selection and mismatch of font families (how many serif fonts can one gal handle??) and the game toolbar overlapped and hid about 1/3 of the gaming screen. 4) human-centered basic design - there were so many features that were obviously needed that Claude missed. For example, the cigarette in one of the game screenshots below was just a dot. And when we had to 'log' the find, we logged it as a cave drawing. And Claude came back and said "false! it's a cigarette!" but provided zero design or caption signaling to tell the user that. Very obvious gap if you're a human user of it, not obvious if you're an AI designing it. The user is still very much the owner of UI QA. 5) human-centered delightful design - it will do the thing you ask. will it add extra levels or a fun mascot or buttons to make things more engaging or whimsical captions to drive the game forward? eh. the game was timed and we asked it to add a 'health' meter, which it called 'stamina', but because it set it up as a timer as well (and not health points), the game had two timers running at once. 6) logic - we had to label the objects we found in the cave as one of four categories - authentic, anachronism, forgery, or 'natural/modern'. Modern is also anachronistic so that made no sense. Also, we labeled one object as 'natural' and the system came back and said "nope that's actually a ghost!" and marked our answer as wrong...but ghost wasn't even one of the four options. 7) iteration and tradeoffs - I want my AI systems to consider tradeoffs. If we're dealing with a massive design overhaul, clearly I'm willing to answer 10 questions about it. If we're changing one font, I'm okay answering 0-2 questions. We asked for a kid-friendly version of the game; Claude asked 0 questions and built for a user that was too young. It felt like it lacked reasoning/empathy or like it didn't make the right tradeoffs. 8) token usage - mine was fine but saw several posts saying it killed theirs, so I want to call that out as a consideration and something to check on with '/usage' inside of Claude Code. So there we go. It didn't replace a PRD process or the deep iteration I do on product building but it made me slightly less likely to recommend to non-engineers to look into Lovable as a design prototyping buddy. 'Claude the Everything Machine' is getting more vertically integrated every day.

Source: Allie K. Miller (X) | 2026-04-20