Ten months ago, I was on a call with a client and I watched them pull up a Replit prototype and say, "We vibe coded this yesterday, can you make your designs more like this?"

My heart sunk.

I didn't know what Replit was or what they were talking about, but there it was. AI made them a prototype overnight.

That was a huge wake-up call for me because I realized that my role was going to change a whole lot in the next few months, and I needed to learn these new tools fast. Not just to stay relevant, but to be able to understand what they could and couldn't do and be able to have conversations like this more effectively.

It took some trial and error, but what I found worked best for me was giving myself the space to try new tools, being inspired by other people, being able to stress test them with real client work, and slowly iterating and practicing to build more mature and consistent workflows.

This article covers my systematic approach to learning AI tools that I used eight months ago and I still use today to find new tools and integrate them into my system. It also includes the tips that I would give someone who's just starting out on how they can figure out how to integrate tools like this into their workflows.

Stage 1: Open-Ended Exploration and Practice

I started out exploring ChatGPT, trying to understand how I could use it in my design process. I started vibe coding with tools like Replit and Lovable. It was a process of spending nights and weekends or whenever I had free time trying out a prompt, evaluating the results, making changes, and seeing what gave me different or better results.

Learn with Objectives

The biggest tip I have for anyone starting out is to give yourself a specific objective or a task to complete.

When I started out learning how to vibe code, what I really wanted to make was an AI project management tool that could turn audio transcriptions into tasks, deadlines, and projects that I could use, kind of like a super-powered AI Asana. What was great about this was I had a very clear direction without the pressure of performance if I was trying to use this for work.

This goal pushed me out of my comfort zone and pushed me to experiment, solve problems, and try things that were uncomfortable. I had to learn about:

• Backend development and how to set up a database that could drive projects and tasks.

• LLM APIs that could transcribe my audio and generate tasks from that.

• How AI can write JSON files and use that as a way to bring data into the database.

• Prompting techniques to achieve my specific goals.

I also learned a lot about prompting. Because I had a plan and a vision and a design for what I wanted to achieve, it helped me practice prompting to try to get that specific goal, and it led me to look up resources and learn more about how to prompt more effectively. I learned that starting with clear goals and a product requirements document can be a helpful starting point for vibe coding, and that led me to practicing using Claude to write a PRD, and it led to me learning about Tailwind CSS and the Shadcn and reading their documentation.

I never actually ended up making a final version of that project management tool that I wanted. But even though I ultimately failed to build what I wanted, having that clear goal and a path to follow gave me direction and motivation to get through the most uncomfortable part of the learning process.

Learn from People Ahead of You

Another huge learning pillar for me was looking at people who were further along in their journey than I was and learning from them.

I found communities on Linkedin and Twitter, be they designers, solopreneurs, indie hackers, or whoever, the people who were at the forefront and pushing the boundaries of what you could do with these tools. People who had been using Cursor way before everyone else, early adopters and designers working at some of these pioneering AI companies like Perplexity or Anthropic who were actually bringing these things to life.

I could see what they were talking about, what they were working on, what they were making, and the tools they were talking about. All of that gave me inspiration for things that I wanted to try and I wanted to learn from. I took their tweets, saved them to Mymind, and tagged them so I could come back to them later. Anytime I saw a tool I wanted to use, I put it in a Linear task to come back to and try when I could.

Because I don't really have coworkers (I'm a freelancer) I didn't have a lot of people in my near vicinity that I could learn from, so I really had to seek those people out. It was a huge help for me in just seeing what was possible and what was out there.

Document Obsessively Throughout the Process

I worked really hard to document things during the learning process. If I found a prompt that worked, I would save it to Notion or Mymind. If I found a resource I wanted to keep, I also saved it to Mymind. When I was working on vibe coding my project management software, I would record devlogs every time I worked on it, trying to capture what I was learning, what I tried, what was working and what didn't work, and what I wanted to try next time. And like I said, anytime I found a new tool I wanted to try, it went into a Linear task for me to come back to.

All of this documenting helped me collect things that I was curious about. So when I had free time, I knew exactly where to go to pick something to explore or a new tool to try.

Documenting devlogs, prompts, and what I was learning forced me to formalize everything I was learning and say it out loud, put it into words. It forced me to reflect on what worked and what didn't, whether I liked the tool or not, as opposed to just going off of gut feelings.

All of this really gave me a handle on the basics and made me comfortable with playing around with these tools. And probably most importantly, it taught me what I didn't know and what I could learn from other people.

Stage 2: Stress Testing in Real Work

The next phase was actually bringing what I was learning out of the lab setting and into the real world.

This was a performance test of what I had been trying. These tools were interesting, but how do they actually stand up to a real project or a real work environment or a real client and their ideas and feedback?

Identifying Opportunities

Now that I was familiar with these tools, I was starting to see different areas where they might fit into day-to-day work:

• Using Visual Electric to generate placeholder images and Claude to draft copy for websites.

• Creating a custom GPT trained on software information for a client struggling with customer support.

• Making code prototypes for AI SaaS clients to prototype with real APIs, real data, and more complicated interactions than Figma could achieve.

Rather than revamping my process overnight, I picked these specific ways to bring AI in to the work I was already doing to test its effectiveness and test my clients' reactions.

Gathering Feedback

The good news is some of this really clicked. The bad news is a lot of it did not.

There were definitely some successes:

• One client loved their custom GPT and started adopting it company-wide.

• Visual Electric and Claude eliminated the need for Lorem Ipsum.

• Claude helped me write code for email templates.

But some of these things didn't pass the test:

• Developers said code prototype interactions were nice to see, but the code wasn't very useful.

• Clients were confused why code prototypes didn't look like their existing software.

• Clients got caught up in smaller details where the AI had made mistakes.

• I couldn't work as fast in code as I could in Figma when responding to client ideas in real-time.

• 
  

The Feedback Loop

Despite some negative feedback, this was validation. This was the real-world test or the alpha launch of my new skill set, and it taught me a lot. I felt validated that these tools are useful. They work and they're powerful, and they do have a place in my workflow. And it definitely reduced my fear and made me feel a lot more confident. The world was changing, but I had a place in it. I just had to keep learning.

I used this feedback to identify the ways I could improve my use of these tools and improve my skills and seek out new tools that might be a better fit. Could I make more realistic prototypes? Could I prompt better to avoid issues? Is there a way that my vibe coding work could connect back to my work in Figma? That sort of thing.

So I took all this feedback back to the drawing board. Same as before, I was recording the whole process. If something worked, I saved it for later. If I had an idea, I wrote it down to come back to it. And I was constantly bringing these ideas, skills, and tools back to my day-to-day work to collect feedback. I was treating my skills like a design project: rapidly iterating, gathering data, making changes based on feedback, and improving.

Stage 3: Mature and Repeatable Workflows

Eventually, the more times the tool went through this loop and proved its utility, the more it made its way into a consistent workflow that I could use constantly. You can think of this step as systems building or infrastructure. Basically taking everything I learned up to this point, taking the things that you know work, and laying the foundation so I could repeat that really consistently.

Don't Make Me Think

The biggest goal of this stage is taking some of the cognitive load off of my plate to make decisions on what tool to use, how to use it, how to prompt it, and all of that stuff.

Some examples of systems I set up for myself:

• Create a prompt library from all the prompts I saved throughout the learning process.

• Set up custom GPTs or Claude projects to have specialized chats with context and instructions built in.

• Explore system prompts or project-level prompts in vibe coding tools to reduce effort and errors, like "Always add animated hover states" or "always add mobile responsiveness."

• Build cross-tool workflows, like pairing a custom GPT trained to be a CTO with a vibe coding tool to review and plan code changes, or using Granola transcriptions to build context for Claude chats.

And this isn't even getting into the whole world of agents and automations, which doesn't play a huge role in my day to day work (yet…?)

Internalizing Prompting Fundamentals

Part of using these tools consistently in my workflow is building up a really strong understanding of the basics of prompting, context, and managing errors. These are true of any tools that you are using.

For example, I learned that:

• Vibe coding tools can get stuck on specific errors, so it's often faster to start a new chat or project.

• Providing more specific and useful context will drastically improve results (through meeting recordings, Google Drive integration, etc.).

• Different models are best at different tasks, reducing the need to constantly switch tools.

• AI defaults to agreeing with me, so I need to explicitly ask for disagreement or negative feedback when brainstorming.

• Prompt output varies significantly, so important tasks require multiple retries across new chats and different models.

• AI can be confidently wrong, so I always proofread and edit generated content.

The point is I'm not just learning how to use a specific tool. I'm building up skills and knowledge and fundamentals and kind of learning the language of this whole new AI world. So I haven't just learned these tools I've already seen, but now I'm better equipped to learn whatever new tool comes out next week.

Wrap Up

So that's the playbook that's really worked for me for learning these new tools and becoming comfortable with them.

Like I said, ten months ago, all of this was foreign to me. I would play around with ChatGPT here and there, but I didn't see it as much more than just a toy. But through guided learning and some trial and error, I was able to push myself outside of my comfort zone, learn new skills, build confidence with these tools, find ways to fit them into real workflows, and ultimately really diminish the fear that I'd been feeling.

I've learned a lot, and I use AI every day, but I'm still learning new tools. And that's a really exciting thing. I'm constantly beginning this journey from scratch, and I still spend time every single week trying to explore new tools, new ways of working, and new ideas to try with AI.

You've learned Figma, you've learned Framer, you've learned Photoshop. You've learned a million other design tools. This process is the same. It's frustrating, scary, and uncomfortable. But there's a way to move forward, and there's really cool stuff on the other side.