Skip to main content
All CollectionsHostinger Horizons
Hostinger Horizons: Best communication and prompt practices

Hostinger Horizons: Best communication and prompt practices

Learn more how to prompt efficiently and get the best results out of the communication with Horizons AI

Updated over a week ago

When it comes to communicating and sending messages to the chatbox inside Horizons, there are some specific communication practices which can generate better results with your tool 🚀

Avoid sending large prompt messages to Horizons requesting multiple changes at once, as this stops the tool from the best performance 🚀

You can find those prompt action practices below:

  • The first message is more high-level and includes what exactly you want to create. It should be around 2 to 5 sentences.

  • The second, third, and next messages should be for iterating your web application by one action at a time.

  • After that, I'd recommend requesting 2-5 sentences as actions inside one prompt per time.

Additionally, don't communicate with the Horizons chatbox as you would communicate with someone from our support or via usual text messages, avoid messages such as "I don't like the result" or "It's giving errors" and instead request a specific action "Fix this [specific situation]".

You can find some additional communication practices that could be helpful:

  • Horizons tool generated results work better when prompting using AI generated text (like Gemini or ChatGPT).

  • You can also speak to Horizons and upload images—for example, provide a sketch of your app or a screenshot of something to fix.

  • If you face challenges, take a screenshot of the situation you face and send it in Horizons chat asking for a fix so that the AI itself would fix the issue

In general, if you face errors, the best practice to fix errors is to click on "Revert to this message" button to the message in prior to the one you started facing errors 💡

The more times you requested Horizons to fix the errors, the more it may have created additional errors inside the tool. That's why the best practice is to revert to the initial message before the error stated to happen.

Did this answer your question?