How This Thought Began
I recently had an interesting conversation with a Reddit user who was asking for help on how to become a software developer. They framed the question more along the lines of: "How do I build with AI and be sure that I'm doing it right?"
It's even a question seasoned devs ask as they hurriedly click 'enter' on their agent's response.
"Is this code good?"
Although public consensus is that models are rapidly increasing in inference, precision, and utility, there is also the shared understanding that you have to know how to use the tool to get the best results.
With some minor React and CSS experience, they felt equipped for frontend development, but didn't have any idea (or confidence) about the backend.
While Claude and Replit and other platforms are marketed as tools for anyone, if you spend any time on Reddit like me (and presumably this user), there are countless stories of things going awry: DBs deleted, missing logic, leaked environment variables. The list goes on and on.
So, how do you know if the code is good? Well, it comes dow to asking: how do you learn?
The Idea
By now, it's widely accepted there are different learning styles. Everyone learns differently, and what's effective for some isn't effective for others.
That's why this Reddit thread was so interesting. Unfortunately, I can only answer based on my own experience, and that's to build! You start off learning the primitives and the very basics, and you grow from there. Even in the age of AI, this is still how I'd recommend starting.
But, what I couldn't help the user accomplish was deciding how to start. That's always the challenge: a million ideas, they all sound great, but how?
In between our years of shared (or not-so-shared experience), the answer came from in between the lines: "why don't you build a system to help you learn?"
Prompt Claude and describe what you want to achieve, what you want to learn, and how you envision the process. Even if you can't envision it, you can still ask Claude for starting points. Replace Claude with your LLM of choice (but I have a personal preference towards Claude Code), and start with your own bespoke method for learning.
"I want to learn backend development 101 for node.js. Create me a game that shows code and let's me fill in the blanks. Let it get progressively more challenging as I continue. Limit to only the basics!"
Thinking Out loud
Here's some ideas you can use:
- Explain It Like I’m 12 Mode – LLM rewrites any concept at progressively simpler levels until it clicks.
- Micro-Win Challenges – 5-minute exercises that reinforce just one idea before moving on.
- Fill-in-the-Gap Learning – LLM gives partial code or partial explanations; you complete the missing pieces.
- Step-by-Step Debug Guide – Instead of fixing code, the LLM asks guiding questions so you find the bug.
- Visual Flow Builder – LLM generates simple step-by-step flow diagrams you reconstruct from memory.
There's an unlimited amount of tools that you can build for your own gain, which is the greatest thing about building with AI. There's not a single right answer, apart from just build that I can give you. We're in the age of ephemeral apps, and it's never been easier to build something lightweight and simple to use.
What you build is up to you, and how you learn from it is equally in your own hands.
I'm a visual learner
Full disclosure: I'm a huge advocate for one very specific platform for learning and using. It changed my life, and if you follow me, I'm going to write about it more!
If you want something that focuses purely on the backend that you can use in the real world, with real applications and users in mind (but also helps you learn), you can build with Xano.
It's an agentic supported backend development platform that lets you visualize your code. Having pivoted from the enterprise nocode space to becoming a fully self-contained, auto-scaling backend platform that supports MCP development, it's my go-to platform for hosting my web applications.
It's where I solidified my understanding of logic and backend development, without the need for the DevOps and the parts that make you question "is my code good? Am I doing it right?"
There's more to the platform than what I've said, but the fact of the matter is, it doesn't matter how you learn or where you start, you just have to build.
Learn!!
Whether you're starting out, building ephemeral applications, SaaS applications at scale, or anything in between, the objective should always be to learn.
Oftentimes we get lost in the memes of Reddit, but incidentally, the comment section is where the gold is at. Shared perspectives, learned experiences, praise and critique: you can learn a lot from interacting with these users. Some things might appear obvious, but it takes a conversation to uncover.
To reiterate: there's not a single right way to learn, but you can learn a lot from others and their mistakes, their successes, and their strategies; you can use these lessons to build your own tools to help move your forward.
So the question becomes, how do you best learn? Have you built anything to help you?
Top comments (0)