It is easy to get fed up with all of this
Claude Code, OpenAI, ChatGPT, Codex, Grok, Kimi, DeepSeek, n8n… And then there are skills, plugins, MCP.
If you want to start using AI but have no idea where or how to begin, this post is for you. Honestly, if I were starting today, I would probably be fuming. So many tools, so many tech bros and AI influencers. Everyone uses something different, and everyone says their choice works better. They are all competing to tell you what to pick.
You could spend an entire evening comparing those names and still have no idea what you actually opened your computer to do. So I want to show you one ordinary task: turning a few loose sentences into a to-do list. You will be able to judge for yourself whether AI helped.
Let’s actually open the chat
It could be something as ordinary as organising a picnic. Ola is bringing a blanket, you are buying bread, the weather needs checking. Nobody has volunteered for drinks yet. A short list would come in handy, before you get there and discover you have plenty to eat and nothing to drink.
In a moment you will get a fictional note to try this with. Open an AI chat in your browser and start a new conversation. If you have not used one yet, there are links to ChatGPT and Claude below this paragraph. You will need the usual box for typing a message. We can leave files, search and extra tools for another time.
Tell it what you want to do with the note
Write what would be useful to you, then give it the material to work from. Here we want a short list of decisions, tasks and questions. If you have come across the word “prompt”, that is what it means here: the instruction you type into the chat. Just saying “help me with the picnic” leaves quite a bit to guess.
OpenAI’s documentation describes separating instructions from context, and Anthropic recommends being clear about the desired output. Our message also asks AI not to fill in decisions that have not been made. The drinks should stay an open question until someone actually agrees to bring them. Copy everything below, including the note, and send it as one message.
It looks reasonable. Let’s check the note
Below is an editorial example of what a useful list might look like. It is not a recorded response from a particular model. Your chat may use different words, add headings or arrange the points differently. Start by checking whether it kept the information from the note.
The picnic should still be in the park on Saturday at 12:00. Ola volunteered to bring a blanket; the person writing “I” is buying bread. Nobody has taken on drinks. And the weather? We know it needs checking on Friday. Who will do it is still undecided. That missing name is useful: it reminds you what you need to ask about.
If your list also says the bread must be bought on Friday, go back to the note. Friday referred to the weather. A shopping deadline has been added. It is easy to notice in a picnic note; in a long document you might miss it. That is why OpenAI recommends access to original notes when reviewing a summary. Anthropic also notes that even advanced models can produce text inconsistent with their context.
What if Ola got the drinks as well?
Suppose AI gave Ola the drinks even though she only volunteered to bring a blanket. Go back to the chat and say what does not match. You can also say where the information should go instead. The message below makes that correction. If your list is right, feel free to skip it.
Read the list again after the correction. “Sorry about the mistake” does not tell you whether Ola has actually been let off drinks duty. Check the result itself, because that is what you will be using later.
Now take something of your own
Once the list matches the note, copy it into your own notes. You can add a name beside the drinks when someone volunteers. You have organised the material and can see what still needs deciding.
Now think of a small task you actually need to do. Perhaps there is an overlong email somewhere, or a few sentences with tasks buried in them. Choose material you can share with the service you use, then explain what you want to get from it. You know your own text, so it will be easier to tell whether the answer helps or gives you more work.
And when you feel like making something visual, explore our projects. You can change a poster or build a small website. There is further reading on the AI step by step page. Pick something that helps you do a thing you want to do.
Sources and further reading
The sources behind this article. Instructions and example scenarios are suggestions from our editorial team.
- OpenAI: Prompt engineering (opens in a new tab)developers.openai.com
- Anthropic: Prompting best practices (opens in a new tab)platform.claude.com
- OpenAI: Safety best practices — human review (opens in a new tab)developers.openai.com
- Anthropic: Reduce hallucinations (opens in a new tab)platform.claude.com