Generate, iterate, refine

Before today, when you used our pack generation or Magic Import tools, the process was a simple, one-way street: you gave us a prompt, and we immediately generated content for you.

If the result wasn’t quite what you were looking for, you had to go back, tweak your prompt, and try again. It often felt like you were trying to guess the right magic words, repeating the process until you finally got the content right.

We’re excited to share that we’ve updated this workflow to make it significantly more intuitive and useful.

Ask for adjustments and iterate to perfection

The big change is the addition of an intermediary, conversational step between your prompt and the final content creation.

Now, you give us a prompt, and we’ll generate an initial content suggestion, along with an ongoing chat window. This is your chance to talk to the AI and ask for adjustments. You can:

• Ask for more or less content.
• Request additions or deletions.
• Correct the format or content based on the initial preview.
• Ask the AI to change the focus or difficulty.

You can now iterate on the suggested content with our AI until it’s just right. Once the content is exactly how you want it, you can take the final action, whether that’s creating a new pack or adding words to an existing one.

We believe this conversational approach makes Knoword’s content generation tools significantly more powerful and useful.

A real-world example

Sometimes, the best way to show a change is with an example.

Let’s say my partner, who loves horror movies, wants to generate a pack connecting famous horror movie villains to the movie they’re from.

1. She inputs her initial prompt into the Generate a pack tool: “quiz me on horror movie villains based on the movie they’re from”
2. The resulting pack is almost perfect, but the AI defaults to using a description of the villain as the clue instead of the movie title itself. It missed the mark on the format.

With the old system, she would have had to delete the pack, rework her prompt, and generate another pack.

How it works now

1. She enters the same initial prompt: “quiz me on horror movie villains based on the movie they’re from”
2. She sees a preview of the generated pack and instantly realizes the clues are wrong.
3. She types a subsequent message into the new chat window: “the clues should be the movie that the character is from”
4. The AI responds with an updated suggestion, correctly mapping the horror movie villains to their movie titles.
5. She could even take this a step further and say, “include the year that the movie is from,” and just like that, the year is added after each movie title.

Try it out today

We hope you enjoy this update and find that generating high-quality content for your packs is now easier and more reliable than ever!

If you have any feedback on the new conversational flow, or if the system isn’t working exactly as you expect, please don’t hesitate to send us a message.