Back in Figma, without the crutch
I spent a lot of time in Figma this month without depending on AI for the actual design work, and I really enjoyed it. I explored Figma Draw, Motion, and advanced prototyping with variables and conditions. I also tried Weave, but I see that more as AI experimentation than normal design work.
After more than three and a half years in design, I still feel that taste, judgment, and attention to detail matter a lot. At the same time, I'm learning HTML, CSS, JavaScript, and other basics because I want to understand how the things I design actually work.
The tools I'm actually using for code
Right now, I'm using Claude, Codex, DeepSeek, and Devin for coding work. Devin surprised me because it removes some of the small decisions around models and tools. I like that. Choosing between models is also mental work, and I would rather save that energy for the actual problem.
Model selection is a product problem
Working on a product recently made me realize that choosing the right AI model is becoming a product problem itself. Most people should not need to know which model is faster, cheaper, smarter, or better at reasoning. A good product should understand the task and make most of those choices in the background.
Advanced users can still have control, but normal users should not need to understand the whole AI market just to get something done.
The line between design and code is getting thinner
Designers are learning more about code, and developers now have better tools for working with design. I don't think every designer needs to become a full engineer, but understanding how products are built can make you a better designer. That is one of the main reasons I'm learning code now.
I don't need AI to pretend it's human
I'm not interested in AI trying too hard to feel human. I don't need fake laughs, forced emotional reactions, or casual filler. I know I'm talking to a computer, and I'm fine with that. I want AI to be powerful, clear, useful, and good at doing the work I ask it to do.
Research tool, not a replacement for thinking
I still use AI a lot for deep research, brainstorming, outlines, diagrams, and thinking through problems. For me, that is where it is most useful. The problem starts when I depend on it too much for the final result. AI writing can still feel too polished, too safe, or too generic. I'm trying to use AI less as a replacement for thinking and more as a tool to make my process faster and smarter.
I've also learned to be careful with how I ask questions. If I tell a model that I like an idea, it often starts agreeing with me too much. Sometimes I don't need agreement. I need someone to tell me what is weak, what I missed, and what could be better.
I probably wrote too many instructions
Over the last month, I created many instructions, skills, and Markdown files for AI. Now I'm starting to think I may have overdone it. Too much context can make things more complex than they need to be.
I'm thinking about deleting most of it and starting again with simpler instructions. Maybe better models need less control, not more.
AI memory is useful but still feels strange
AI memory is still something I'm not fully comfortable with. It is useful when a model remembers an important project or detail, but sometimes it feels forced. When I see a small summary containing months of my work, ideas, plans, and habits, it reminds me how much information I have shared. It does not always feel helpful. Sometimes it just feels strange.
I do like being able to search old chats when I need something. That feels different because I am asking for it. I know why the model is looking into the past, and I have more control over the process.
Backend yes, visual design not yet
I enjoy using AI much more for backend work than for visual design. With backend tasks, I can explain what needs to happen, test it, and see if it works. With design, I already have an idea of how something should feel, and AI often misses the small details. After enough prompting, I usually open VS Code and fix things myself.
That frustration is actually helping me learn. I change something, then ask AI what it does, why it works, what could break, or what would be a better way to build it. I'm not trying to become a developer overnight. I just want to understand what happens behind the interface, because I think that will make me a better designer.