Hey y’all, my name is Sam Solomon. I’m currently the product design lead at Salesloft. I’ve been with the company about nine years and have been building software for about 15 now.
I keep getting these questions about AI—incorporating AI into workflows, how do designers start using AI, how do product orgs start shipping things with AI? We have these questions internally too. I built a couple of tutorial videos that the team caught onto, and it occurred to me that maybe I should just start creating videos showing what my workflow is.
The thing I’ve been trying to do with most designers is get them past the plateaus. There are these different ridges you get to, and then there’s a really big wall that’s hard to get over. One of the biggest is honestly if you’re at ground zero. You don’t have to have any coding experience here at all. What I’m going to do is take you through the process to get you set up to the point where you know what questions to ask. From there, we can start building applications.
I’m going to try and split this up into a bunch of different videos and edit them so we’re not just waiting for Claude or Codex to finish doing things. Hopefully this is a pretty information-dense tutorial.
What we’re going to do right now is get you set up. I’ve got a bunch of stuff installed on this personal computer—I’m going to show you what you need. If you’re on a Windows computer, it’s probably going to be a little different. Macs have terminals; Windows has PowerShell. A lot of the stuff in theory is the same.
We’re going to get a Claude or ChatGPT account, look at terminals, install some software and use CLIs for coding. CLI stands for command-line interface—an application that runs inside a terminal. You’ve probably seen the Claude Code desktop app or Cursor, or things like Lovable.
My personal preference is to use CLIs. If this is new to you, it’s going to show you that the terminal isn’t as scary as you think, and it’ll give you a better sense of how file systems work.
If you’re on a Mac, you can type “terminal” and you’ll get the built-in one. That’s fine to use if you don’t want to install anything else, but I think there are better options.
The one most engineers love is called Ghostty (with two T’s). It’s an open-source project by Mitchell Hashimoto—a famous engineer, also the founder of HashiCorp. Ghostty is fast, configurable and easy to split into swim lanes. That’s important because when you get used to this, you’ll want to start running multiple agents at a time. I personally don’t usually run more than four; in reality I’m probably running two.
The other option is my favorite, and a little more controversial—Warp. Warp is nice because it has a true text field. The argument against Warp is it sends telemetry back to the company. If you’re using this for work, I probably wouldn’t advise it, but for a personal project it’s great. For this course I’ll be using Warp.
The next thing is making sure you have either Claude or ChatGPT. I think for both of these to code you need a $20 account. Claude is by Anthropic; ChatGPT is by OpenAI, and Codex is what they call ChatGPT’s coding agent.
We’re going to install the CLI here. If you don’t like it, both have desktop apps. There’s also Cursor and a lot of options out there. If you don’t like the command line, I ask you to sit with me and try it first.
Before we install the agent, we need something called Homebrew. Go to brew.sh. Homebrew gives you the ability to install software on your Mac for developing software that most Macs don’t ship with.
Copy the install command from brew.sh and paste it into your terminal. It’ll ask for your Mac password. When it runs, that’s pretty much it.
Now that we have Homebrew, we can install another critical piece of software: Node. JavaScript is an incredibly common, powerful language that runs on the web. Node is what allows JavaScript to run anywhere—locally, outside the web.
Run brew install node. With Node we get something really important called NPM—Node Package Manager. You don’t need to know a ton about it, but that’s really the main reason we’re installing this.
Now we can install Claude or Codex. If you signed up for Claude, find the install command in their docs and run it in your terminal. After it installs, just type claude and hit enter—the CLI opens. It’ll ask whether you trust this folder. We’ll talk about navigating later.
To exit the CLI, hit Ctrl+C twice. That’ll bring you back to your regular terminal.
There are two other CLIs to look at. If you use ChatGPT, install Codex instead. Same install pattern via NPM, then type codex to open it.
A third option that’s actually one of my favorites—but you can’t use it with Claude—is OpenCode. It’s great if you’re interested in OpenAI models, Kimi or any of the other options. Install via NPM, then type opencode. The CLIs we’re installing are essentially harnesses; they all work the same way.
If you’re trying to decide between Claude and Codex—they’re both very good and they go back and forth. Right now, in my opinion, Claude is very expensive. You burn through tokens fast and don’t get as much mileage. GPT-5.5, the latest model you can use with Codex, is really, really good, and you’ll get really far with just the $20 a month plan.
Hopefully now you’ve got everything set up. The last thing I want to do is explain file folders.
A few commands matter here. The first is ls—it lists where you are. If I go back a level to my home folder and run it, I see Applications, Developer, Downloads—the same folders I’d see in Finder.
The next important command is cd—change directory. Running cd developer moves me into that folder. Then ls shows all my projects. To go back a level, cd ... To jump all the way back to your home directory, cd ~.
You don’t have to step one folder at a time. In Warp, hit the right arrow for autocomplete (Tab in Ghostty or the built-in terminal). I can type cd developer/solomon and jump straight in. That’s most of what you need to know about moving around.
The last command we need is mkdir—make directory. Let’s go back to the developer folder and make a new one. I’d keep these project folders lowercase and avoid spaces; if you need a separator, use a dash. Spaces aren’t great in code.
Run mkdir project-project, then cd project-project. Run ls—the folder is empty. Now we’ll start an agent. I’ll run claude, and the little Claude alien comes up.
Why does it matter that we create a folder and install Claude into it? You want to keep things sandboxed. This folder represents the project you’re going to build—the to-do list, the bookmark app, whatever. All the code is associated with this one folder.
The reason matters is that as the project grows, the agent isn’t always going to know where to look. If we started it at the developer level, it might go searching through every project in the list. That’s a waste of time, a waste of tokens and gives the agent a lot of context that isn’t useful. We’ll talk more about managing context later.
At this point you should have a terminal picked out—Terminal, Ghostty or Warp are all great options. You’ve installed Homebrew and Node. And depending on what you picked, you’ve got either Claude, Codex or OpenCode installed.
Hopefully you’ve got the basics. Next episode we start building things. Thanks.