How I Keep Learning AI Without Trying to Keep Up With Everything

I’ve been thinking about 21-year-old me learning how to build web pages by hand, writing HTML in text files, saving work to floppy disks, and using dial-up internet in cyber cafés to understand how the web worked by making things.

Very late 90s / early 2000s energy.

Different era and tools, same approach.

Back then, learning the web meant opening a text file, writing code by hand, saving everything carefully, refreshing the browser, breaking something, fixing it, and slowly understanding how the pieces fit together.

It was mostly curiosity, trial and error, and the faint terror that you had in fact broken the whole company website at 4am because you missed one closing tag in the update.

At the time, the internet was still emerging. Nobody really knew what it would become, what jobs it would create, or how much it would change the way we worked.

That’s part of why the current AI moment feels familiar to me. Different technology and risks, but a similar sense of standing in front of something that’s still forming.

I think about those early web days a lot when people ask me how I keep up with AI, especially as someone experiencing this shift in the middle of a full life of family, business and fun.

We don’t know what AI is going to stabilise into, in the same way that 1999 me didn’t know what building websites would look like by 2009.

So for me, keeping up with AI is not about chasing every update. It’s about building a little learning practice that gives me some agency inside the uncertainty.

I try to keep learning as part of my daily practice. That might be watching a ten-minute video, reading an article, testing a prompt, trying a small build, or planning what I want to experiment with next. I pace the learning to the time and energy I have, but I try to do at least one small thing a day.

For me, that little practice has roughly three parts:

Build: make small things that solve real problems for me, even if they are rough.

Learn: follow enough of the field to understand what is changing, without disappearing into the noise.

Play: test low-stakes or fun ideas, because that is often where the tools start to make sense.

1. Learn enough to ask better questions

You don’t need to understand every technical detail to use AI well, but it helps to know enough about how the tools work to use them with better judgment.

A bit like riding a bike. You don’t need to understand all the physics, but it helps to know how the brakes work, how the gears respond and what happens when you turn too sharply.

For AI, I think that means getting familiar with a few practical ideas:

• what a token is, and why context windows have limits
• how the prompt you give influences the response you get back
• how an AI model is generating a response rather than “knowing” in the human sense
• why adding useful context and source material can improve the answer
• what makes one model better suited to a task than another
• the difference between asking an AI a question and giving an AI agent a task to complete using tools
• what information you should avoid sharing
• where human review is needed before anything is acted on

You don’t need to master all of this before you begin. But understanding these basics helps you ask better questions, choose the right tool for the task, and review the output more carefully.

You can also use AI itself as part of the learning process. Ask it to create a learning plan for your level, explain a concept in plain English, quiz you, compare two ideas, or walk you through an example step by step.

The important thing is to use it as a tutor, not just an answer machine. If your AI tool has a learning or study mode, that can be useful too because it is designed to guide you through ideas rather than simply hand over an answer.

2. Play and test, then focus

At the start, I experimented with lots of different AI models and tools, including open source models.

That was useful because it helped me understand the differences between them. Some are better for reasoning, presentations, spreadsheets, coding, research, summarising or helping with a messy idea.

But at some point, I think it helps to stop trying everything and go deeper into the tools that fit the kind of work you are actually doing and that suit you.

For me, that has meant spending more time with OpenAI tools, especially ChatGPT and Codex, because they suit the experiments I am building.

That doesn’t mean everyone should pick the same tools. It just means you don’t have to chase every release to keep learning. You can keep an eye on the wider field while building depth in the tools that are most useful for your own work.

3. Match the model to the task

Part of learning AI is learning which tool to use for which task.

You don’t always need the most powerful model for everything.

For example, I probably don’t need to ask a high-reasoning model how long to air fry chicken legs. That’s probably not the best use of it, and I have in the past accidentally forgotten to switch down a model and hit my rate limit by carrying on a chat with a more powerful AI model. So be aware of the different models available to use and what they are used for.

But if I am asking AI to help reason through a workflow, debug something, structure a complex idea or understand why a build failed, then using a stronger reasoning model makes more sense.

That is part of the learning curve.

Learning how to choose the right level of model and tool for the problem in front of you is really important.

4. Build something tiny

This is probably my biggest advice.

Build something small. Really small.

A web page. A birthday party invite. A tiny calculator. A small thing that helps with one annoying task.

The point is not to build something impressive. The point is to learn by doing.

When something breaks, I try to treat it as feedback: was the instruction clear, was the context strong enough, and was this the right tool for the task?

I’ve found that building small things teaches me much more than trying to absorb every update from the outside. One of the first tiny tools I built was a simple converter that turned Word doc blog drafts into clean plain text, which solved a small but annoying publishing problem for me.

The experiments don’t have to be huge. They just have to be real enough that you learn something from them.

5. Bring your own experience to the tools

Your own background matters more than you might think.

The most useful AI experiments often start with something you already understand well: an industry you have worked in for years, a messy operational process you know inside out, a small business admin problem, a niche interest, or even a personal obsession like Japanese stationery.

That existing knowledge gives you better questions to ask.

For an experienced tradesperson, that might be a calculator to help quote jobs. For a small business owner, it might be a workflow that turns enquiries into next steps. For a stationery nerd, it might be an app that shows all the brands that carry a specific gsm of notebook paper for your fountain pen.

The most interesting AI use cases often come from the knowledge you already have.

Your domain knowledge is not separate from the technology.

It’s what makes the technology useful.

6. Learn from good sources and people

There are a lot of AI courses, posts, threads and tutorials out there. Some are useful and some are more hype than substance.

I find it useful to keep a mix of sources nearby: foundational reads, newer books on AI systems and engineering, official documentation, and research from places like MIT and other universities or labs. Not everyone needs to read textbooks to use AI well. I love reading and it’s part of the way I learn, but the main point is to look for sources that sit underneath the hype cycle and help explain how the tools actually work.

If I was starting again, I’d start with the free resources from the AI companies themselves and other reputable sources before jumping into expensive or overly dramatic paid courses. The official tutorials and docs are often more useful than people expect, especially if you’re trying to understand what the AI models are actually designed to do.

I also like learning from people who are actively working in the field: researchers, engineers, educators and builders who are testing these tools in public. Watching someone explain a concept clearly, demo a workflow, or share what they learned from building something can be far more useful than trying to follow every headline.

The point isn’t to follow everyone. It’s to find a small set of good sources that help you understand what is changing without getting pulled into every “hot take.”

7. Find other people who are learning too

Community matters more than people realise.

It mattered in the early web days, and I think it matters now. Back then, learning often happened through other people after the two-hour lectures: someone next to you in the computer lab showing you how they fixed something, a forum thread, a borrowed book, a half-working example, a person who knew just enough to help you try the next thing.

AI feels similar to me.

You can learn a lot from documentation and tutorials, but being around other people who are experimenting opens up the creative possibilities. Someone else’s strange little build can make you realise what’s possible. A meetup conversation can sharpen an idea. Watching someone demo how they use a tool can teach you more than a dozen abstract posts about productivity.

That’s one of the reasons I’ve found attending meetups valuable and the Codex community so useful. It’s not just about technical answers. It’s about seeing the different ways people are approaching the same tools.

Learning AI doesn’t have to be a solo attempt to keep up with everything. It can be a little practice you build alongside other curious people.

My answer to “how do you keep up?”

So my answer to “how do you keep up?” is probably this:

I don’t.

I try to stay curious enough to keep building.

There is no one right way to learn AI, and there is no one right way to use it. Your learning practice will be shaped by your interests, your work, your constraints, your curiosity and the problems you want to solve.

For me, that means understanding the basics, trying small things, using my own experience, learning from good sources, finding other people who are experimenting, and keeping creativity in the loop.

Most importantly, jump in, have a play and keep learning in the little moments.