The Simple Way To Learn AI From Scratch Without Getting Overwhelmed

Navigating learning AI today can be daunting, in part because so much of it is either outdated or theory you won't ever apply in practice. It's better to just ignore most of the noise and concentrate on the part that is small but which is actually going to be useful and which is likely to be useful for years to come. That's broken down into three levels that each build upon the previous, from the basics to a system that gets smarter the more you use it.

Level One: Pick One Assistant and Go Deep

Why Choosing One Is Enough

The gap between the top AI assistants is “negligible” to the common user. In addition, they have similar core features, so the skills that you develop on one transfer to the other. You don't have to try and compare a million choices. For most, the three realistic options are Claude, ChatGPT, and Google Gemini. Other tools focus on searching products based on different models or remain a long way behind the leaders.

How To Choose Between Them

There are three easy rules to make the decision. First, opt for a paid plan. There is a significant difference between the free version and the paid version, and if your workplace provides you with a paid version of one assistant, learn that one. Secondly, match the assistant to your work. ChatGPT is the most popular, has the most tutorials, and is good at web search and research. Though code is hard to do, even for non-technical people, charts and data analysis are built on code in the background, and Claude excels at writing, designing, and coding. If you work with text, images, audio, and video at the same time, or use Google Workspace extensively, then Gemini will be a great fit for you. Thirdly, trust your gut. Every assistant is unique, and the more you use an assistant you like, the more you will use it, and the more skilled you will become at using it. It's possible to switch later, as most will allow you to import your memory.

Change The Default Setting

Companies typically use their lowest, least expensive model as their default. When it comes to real work, always go for the most capable model that you have. These more powerful models dissect requests, chart a course, and uncover information you wouldn't have thought to include.

Level Two: Give The AI Good Context, Not Perfect Prompts

Why Context Beats Clever Wording

Fortunately, the language of the prompt is not as important as it used to be for some of the models available today. It's important to provide the assistant with a clear goal and context. Consider offering a suggestion for your boss's dining out. Take 10 minutes to tell your boss what you believe they enjoy, or give the assistant a list of restaurants they've always enjoyed in the past, and the boss will take the rest of the time to figure it out. The second strategy almost always works.

A Simple Formula To Remember

There is only one framework: outcome plus context. If you desire a workout plan, say it. If you're looking for a workout plan, say it. Instead of a long text on your experience, equipment, and schedule, copy and paste an article detailing a training schedule you like, and write a short request like, "I want to build a four-day muscle growth plan with a 45-minute per day training routine based on this article." The assistant can assume that the context has a more refined role and structure than you could possibly come up with.

Three Ways To Find The Right Context

It is advisable to name a relevant framework rather than give an explanation of it. The pyramid principle is more information that you can give an assistant to rewrite something than a paragraph to describe it, and you could even ask the assistant to give you ideas about what the frameworks for your goal might be. Second, provide authentic exemplars of good work. If you place the final x number of approved weekly updates and your raw notes, the assistant will copy the format, length, and tone without being instructed. Thirdly, connect your tools. Your best context is frequently in e-mail, documents, messaging applications, or notes, and having the assistant look at them directly saves downloading and uploading files.

Save Your Context in Projects

If the work is repeatable, most assistants will have a feature that they call "projects" or something similar, which is a permanent home for a workstream. A project is made up of three parts: instructions with rules that are always in effect, knowledge files containing your reference material, and memory that automatically updates as things change. Imagine you have your limits—45 minutes of exercise at home—and you have your training plan. Your workout project could include both of these. Later, when you ask what to do with only a quarter of an hour, it comes up with an answer, and when you talk about injury, it recalls the information and does not go through those exercises. Markdown files tend to be easier and cheaper for the assistant to read than PDFs, and you can ask the assistant to convert files.

Level Three: Build A System That Connects Everything

What Projects Cannot Do

The projects remain independent of each other; you can't have your fitness program data in another project simply because they're obviously linked. The AI system fills that gap in two ways: it pulls context across projects and finds patterns that would not be spotted by individual projects and increases as it learns from your feedback over time.

Real Examples Of A System At Work

If your health checkup, supplements, and workout plan are all integrated into one connected system, the assistant can see that your most recent health checkup has borderline high cholesterol and your workout plan doesn't offer any cardio days, then recommend that you add cardio to your rest days. It would have been impossible to connect these dots without people doing the copying. One is a writing feedback loop, for instance. You edit an AI draft, and then you ask the assistant about how it differs from your first draft; it calculates the differences you've made and saves the preferences as a rule so that the next AI draft requires fewer edits.

Choosing A Tool For Your Comfort Level

There are options available that are suitable for beginners as well as those that are quite technical. Some are already integrated into your existing apps, such as email, calendar, documents, and so on, and require minimal configuration; however, they lack the same level of control. Others provide additional control in some setups and are for non-technical users. The most powerful options are fully customizable but will require some comfort with code. The degree of choice in each tool's settings is a good indicator of who the tool is designed for.

My Honest Take On Learning This Way

My initial experience with AI involved a lot of trial and error in tools and finding the right prompt structure, which took me weeks. What the trick really was was to stick with one assistant and to feed it better material than better wording. I used to have a bunch of prompt templates, but pasting in a couple of good examples saved me more time than any of those templates. It also made everything feel like I was just showing a new co-worker what good work looks like in my context.

Where To Start Today

The first level is to select one assistant, the best available, and master it. Level two is understanding the context that works to support learning and saving that in projects so that you don't have to repeat yourself. At level three, these projects are brought together as a system that gets better over time. Keep in mind that most are still at level one, which is okay. Choose one step, practice that step this week, and if it feels natural, then do another step.