A useful way to begin learning AI is to choose a small task whose result you can check. Instead of collecting another long list of tools, take a set of your own notes and turn it into a clear summary. This simple exercise helps you observe what a model understands, what it misses, and which decisions still need your judgment.
1. Define the result before opening a tool
Choose material you understand that contains no private information, such as your notes from a public lesson. Describe the outcome first: a summary of no more than 150 words, three key concepts, and two questions for further exploration. Specify the audience and structure so you have something concrete to evaluate.
Use three acceptance criteria: Does the summary preserve the main ideas? Does it introduce claims absent from the notes? Could a reader understand it without the original material? Fluent writing is useful, but traceable content matters too.
2. Give context, a task, and clear constraints
Try a prompt like this: “These are my learning notes. For a reader new to the topic, write a summary of no more than 150 words, followed by three key concepts and two follow-up questions. Use only the information in the notes. Mark uncertainty rather than adding facts that I have not supplied.” Then add your material.
You do not need a perfect prompt for every possible situation. Make this task specific, examine the result, and adjust it. If the summary is too abstract, ask it to retain an example from the notes. If the priorities are unclear, extract the main claims first and choose which ones to keep yourself.
3. Treat the first answer as a draft
Check the output against the original material, one claim at a time. Identify omissions, misunderstandings, and unsupported inferences. When a number, name, or conclusion needs a citation, open the original source and verify it yourself. A confident tone is not evidence.
Change one variable at a time, such as the audience, length, or format. Keep the before-and-after versions and see whether the change actually solves the problem. This gives you a reusable lesson rather than another unexamined regeneration.
4. Keep the process, not just the answer
Create a short practice log: the task, source material, prompt, mistakes, corrections, and total time spent. Record the tool or model name as well so future comparisons have useful context.
Even if the result saves no time, the exercise can teach you something. Perhaps the source notes needed better organization, a particular judgment required human review, or the task was a poor fit for the model. Documenting these limits is part of learning.
5. Expand by one step
Try the same method on a second set of notes and check whether it still works. Once the process is reliable, add a single step, such as turning the verified summary into revision cards. Keep each result small enough to inspect.
This is the kind of exploration Let’s AGI aims to document: start with a real question, run a small experiment, check the result, and leave a useful record. The horizon of AGI may be distant; your first practical step can be small.