This one is about how I ended up going to the terminal. Thinking about it is very interesting, because I would never think that would be something I would do. When I first started using Lovable, and then Cursor, I actually did think Cursor would be the last app I would be using. Then I went back to VS Code, and it revived all my years of small little projects here and there. PHP in the beginning, and later on very simple small scripts. I never actually thought VS Code would be something I would be using again. And I used it for a very small amount of time. From Lovable to Cursor, from Cursor to VS Code, I think that was the span of two months, three months tops. It was me trying out these new tools, but not feeling fully inside the tools, the way I feel now when I'm using the terminal. The folders make no difference I felt confident with the terminal from the very beginning, even though I didn't have the visual part. And I think this is the thing. If you are a programmer and you are using AI, it does make sense to use the folders you can see in VS Code, per project and everything. The thing for me is that those folders make no difference in my organisational process, because I don't have that background. For me, what matters is prototyping, using it, and feeling like it can deliver what I want. Having the code in front of me, happening. That is what you have in the terminal, much more clearly and faster than in another tool. Again, Claude Code organised my life The move from VS Code to the terminal was very connected to Claude Code. I managed to have everything ready to go to the App Store inside VS Code. Then I started doing multiple tabs from multiple projects, and things got disorganised. Again, Claude Code organised my life. I had very little understanding of GitHub, but Claude Code organised all the worktrees. Later on I had to clean all those worktrees, because they were local and they caused me issues. But that was much further ahead. At that point, I had no idea what was happening. I was working in a project, and suddenly I would have the idea for another project. I would open another tab and start working, and then I would prompt something in a tab that was meant for another tab. A complete crazy thing. So I watched some YouTube, and some Instagram accounts that I follow. One of them was saying: just move to the terminal. I was also looking at people organising memory, which was something I was starting to feel very clearly with my projects. Even though I had the roadmap, I didn't feel the brainstorming sessions were working well, and I wasn't understanding how to organise my memory. That was an inflection point. It was exactly the moment I moved to the terminal and started using Obsidian. And the whole concept of the second brain started to flood everything. All my feeds were just about second brain. I was always more pro privacy, and having things a little bit more under my control. So Obsidian, for me, was pretty interesting. As someone with the whole Linux background of ricing Linux, obviously the first thing I did was spend three, four days obsessing over the best Obsidian setup. A lot of hours of YouTube videos and guides. And the funny part is, in the end I used the simplest organisation possible: the one from the OpenAI guy, Andrej Karpathy. I strongly advise people to check it out, because it helps you organise your Obsidian into something that works when you start evolving into memory, and the memory issue you are eventually going to have. Run my life So I moved to the terminal. The first thing I did when I opened Claude in the terminal was tell it: run my life. I seriously just said: run my life. Then I thought, what would be the best way for Claude to run my life? You can create commands in Claude. I had watched a guy organising Obsidian, and he had some commands he built. I thought that was a very good idea. So over a couple of weeks I built commands like start my day, end my day, organise what I need to learn, look at my week. And one of the first things I did in the terminal was build my own RPG game. I had a lot of small tasks just falling through the cracks. Doctor's appointments, basic stuff, life stuff. I didn't build the game immediately. I built a command called boring game, to show me all the very boring tasks I had for the day. That eventually moved into another command, called life ops. I used boring game for a couple of weeks. Things moved, because I wanted XP, all those things you can build into a small retention flow. It was pretty fun to have a game to help you out with your everyday life. A website, without WordPress Then I actually built a website for it. Wallfarer. I think it was the second website I did without WordPress or any other platform. I had published websites before, with WordPress and many others. I don't remember all the names. I did a lot of SEO work, so I had to go through all the config, but mainly HTML. I never actually coded coded. Maybe a little bit, many, many years ago, when blogs were a thing and you were working on your Tumblr. Basic, basic stuff. This website took me some time, because it was going to have a login. That led me to learn a lot about which platforms were safer to use. Coming from the Manus fiasco, and everything I learned about not trusting an AI tool, I was building things differently. I looked for a proper platform to run my login. I started to worry about security, obviously. I collected prompts and watched things about cyber security, to have a layer of security on top of my projects and products. I had a very specific story in my mind for Wallfarer. I'm a big fan of science fiction, and I wanted something with The Three-Body Problem. I worked with China for almost eight years, so I wanted a Chinese futuristic look and feel, with some RPG involved. I worked on the design in Claude Design. It was the very beginning of Claude Design. You had tokens, you had to be very careful, and everything took forever. Transferring the design was not straightforward either. Now you can use the Claude extension for a lot of things. At that point you had to download things and do a whole process. The Interrupt website, by comparison, was pretty straightforward, because I knew exactly what I wanted. By then Interrupt was published, and I had started the V2, mainly design, because the mechanics of the app are pretty much set. So I published Wallfarer on Cloudflare, and it was fun. And from that, I felt more confident that I could do other things. The biggest milestone I started moving to my video project. That started back in Cursor. I wanted to do video automation, and I found Remotion, which is open source: skills, rendering, everything you need to do video. My first Facebook Ads campaign used Remotion right away, because the graphics I wanted were static, and it managed. It was the time people were talking about having multiple creatives to test, and I ended up running actual campaigns with the actual work. But I didn't have a platform. I wanted a proper platform, so my husband could do the videos himself, apply presets, apply everything. That was a big, big, big project. It is still ongoing, but now it is pretty much automated. I can run a video from start to end with presets. I upload something, choose presets, and it renders, checks quality, applies the captions, checks the captions. All of that took me a long, long time to learn, because there are layers. There are many layers in learning how to work with AI. There are many layers in learning how to build something that not only works but is safe to use. And then you have the big one: how do I share it? The video project is, for me, the biggest milestone in learning AI. I had to learn everything, from logging to setting up a whole system. Thinking about the logic of the system. Actually having an architecture behind it, because I didn't even know I had to plan for that. And I started to work with Claude in a way where it would be building together with me, but actually te… This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit praximedes.substack.com