Getting Started with AI

The first of a two-part tactical guide on how to use AI tools to build software with no coding experience

Heyo! I’m Cam and I write a periodic newsletter about using AI to build Contxt as a non-technical solo founder. I also answer reader questions about using AI to build their dream software. If that sounds interesting to you, make sure to subscribe below!

Welcome to part one of a two-part series on starting to code with AI:

• Part 1: A Step-by-Step Guide for Aspiring Developers ← This post

• Part 2: Navigating the Full Stack with AI

Let’s get started!

---

In the 16 months since GitHub Copilot X was launched, 92% of software developers have started using AI-powered coding assistants. It makes sense for developers to have this rapid adoption, due to their incredible productivity increases. But where do people like you and I begin?

Even after exploring AI tools extensively, I still feel intimidation, fear, excitement, confusion, curiosity, and trepidation when considering all the things I don’t know how to do yet.

But I no longer have the questions that plagued me for months before using AI: How do I interact with the AI to generate code? What if it gives me the wrong code? How do I debug non-working code? What if I can’t figure it out what my developer friend said “was totally doable”?

These questions are entirely natural and the purpose of this post is to lay them to rest. I’ll outline exactly how to interact with AI, showcase how curiosity is your greatest ally when using AI, and examples of other people using AI. My aim is to demystify AI via play and simple tutorials that anyone can start using today.

Why do you want to build with AI?

I find there’s typically three potential answers to this question:

1. I want to explore and experiment with AI tools, but I don’t know what to build.

2. I want to build a project with AI, but it’s a short-term project with a specific end-date.

3. I want to build a long-term product with consistent feedback and feature development.

All three options are exciting places to be today, but people who selected the 3rd option will benefit most from reading My AI Co-Founder. Launching a project is a worthwhile goal, but it’s different from building a long-term product with customers who depend on your service to function and consistently improve. My AI Co-Founder is for people building the latter.

Your first step is the same regardless of your answer: become familiar with AI tools and the superpowers they give you.

Step 1: Choose your tool

The first step is to choose a tool. The AI space evolves rapidly, so today’s choice may differ from a month from now. For now, I’ve kept it simple with one recommendation:

Claude

Besides industry-leading security, safety, and alignment efforts, Claude is also the best tool for non-technical people building applications without any technical setup. There are 3 reasons for this:

• Claude Sonnet 3.5 - The best model for coding and creative writing. Lots of technical jargon to justify this, but it’s incredibly good and intuitive to use.

• Artifacts - Split-screen mode where Claude renders the code it writes in real-time in a side window. It’s epic and is similar to Devin, but Devin has a waitlist and you can sign up for Claude now.

• Projects - Reserved for paid customers, it’s an intuitive workspace where you can upload documents, spreadsheets, PDFs, or text/code to provide context for Claude’s answers. You can upload entire codebases for Claude to reference when responding to code-related questions.

In Step 2, most examples are experiments shared from Claude because it allows you to share your Artifact via a simple link so others can test and customize your Artifact. Mike Krieger, co-founder and CTO of Instagram, joined Anthropic as their CPO, so I expect the product will continue to significantly improve in utility and ease. They also have about $8 billion in the bank, so they’ll be around for a long time without any worry that they might disappear tomorrow.

Note: there are usage limits for the free tier, so if you hit your limit or want access to Projects, the $20/month subscription will get you there.

Step 2: Build toys

Before starting a project/product, I encourage you to spend a solid 2-3 days exploring AI to acquaint yourself with Claude and its abilities.

This exploration phase is crucial to observe how AI works, take note of the concepts and details of software building you don’t know, and ask questions about those areas to learn more.

Like any toy, the toys you build in this step should be fun and not taken too seriously, things like:

• Build snake

  • What if you changed the size of the dots for the snake to eat?

  • How about turning it into a neon-techno game that changes colors when the snake eats a dot?

• Build a basic calculator

  • What additional styling or features would be fun to have on a calculator?

  • What about making the buttons neon colors?

• Build Flappy Bird

  • Is it playable? Are any changes needed for better performance?

  • What if you changed the Flappy Bird theme to Halloween or 4th of July?

What’s awesome about Claude is that you can type “Build a Flappy Bird artifact” and it’ll build you the game. Is it perfect? Absolutely not. Is it playable? Usually! It’s your job to be the discerning player who catches what can be better.

If you want more ideas, simply ask Claude for some inspiration!

Quick Tip:

To ensure personalized results, always include in your prompt that you are new to coding and want comments in the code and explanations for why Claude approached the problem the way it did.

Step 3: Build a mini-tool

After exploring little toys, it’s time to think about a tool that makes your life 1% easier today.

Observe the mini problems or points of friction in your day. Write them down as you notice them. Imagine what a solution might look like and describe it to Claude. See what it says; have a dialogue with it. Make sure it understands your problem.

Once it does, ask for a prototype. Sure, it’s gonna look shitty, but does it work? Does it help solve your problem at all? If it doesn’t, what should change? How specific do your prompts need to be in order to make the changes you want to see?

Dive into rabbit holes. Ask more questions than you’d be comfortable asking a tutor or friend/colleague. I find it oddly inspiring and liberating because it’s a tangible way to feel true creative agency. I can create toys, tools, and experiments in minutes using languages and frameworks I would’ve needed months to study before.

Here are some examples of the tools I’m talking about. Ask yourself if any of these ideas could be adapted to your situation. Can you add missing functionality or a button to improve them for your use-case?

• List of mini-tools made with Claude Artifacts

• A custom interface for adjusting box shadow properties

• A P&L data input and forecasting tool

• My tool to calculate someone’s age or birthdate when I have different data points

These ideas will get you started. The key here is to start having opinions about how a piece of software should work. If the design doesn’t look right, change it. If the initial prompt didn’t include crucial components, add ‘em! Get curious about what you want, and don’t be afraid to ask as many questions as necessary to learn and build what you want.

Lastly, I’d love to see what you come up with! Feel free to tag me on Twitter if you share any of your creations; I’d love to see and try them out ◡̈

Step 3: Decide next steps

Now that you’ve had your fun, it’s time to get clear on what to do next:

• Did any of your toys inspire you to explore the topic further?

• Do you want to send your toy to people and get feedback?

• Did you have fun and feel satisfied?

There are many blogs and resources online to guide you in finding your startup idea (Y Combinator resources are especially helpful). If you’ve built a toy or tool that you believe is cool and could grow into a legit product, try an AI “Mom Test” prompt (describe your idea in detail and ask the AI to generate 10 questions according to the principles in The Mom Test) to understand if people would be interested in using your tool.

Maybe you had fun exploring and don’t feel a deep need to build something substantial. That’s okay! Now you understand what’s possible, so when you have your next idea, you can return to AI and experiment.

Whether you’re ready to build a full product or just happy to have had fun and learned a bit more about AI, you now know what’s possible. These technologies are improving rapidly and are here to stay, so just remember you have a new superpower and the cape is ready for you whenever you need it ;)

Part 2

Now that you’ve explored current AI capabilities via Claude, it’s time to have an entirely different kind of fun: building something you can use every day and that improves over time as more people use it.

It’s a different process than the one this post outlines, but in my opinion, it’s more meaningful because you’ll need to think deeply about your problem and the product to solve it.

In the next post, I’ll provide an overview of building an app from scratch. We’ll discuss what “full-stack development” means and how you can start your AI full-stack journey today.

If that interests you, subscribe and leave a comment with the questions/topics you’d like to have covered in Part 2 👇