Hands-On AI 4: Is AI Deep Research Worth the Wait?

Last week, we took a look at an agent, Muse, in that case, going and doing things for me, making plans, using a tool, checking the results, and then going around and around and around again.

And if you were there for that episode, then you would have learned about agents and how they look.

The research loop and why I compared it to an agent at the beginning is because it's essentially the same thing.

The deep research tool is that loop, but applied to a specific question.

So first, the agent will go in and will make a plan.

So it will say, hey, this is what I think you're looking for.

And And they will occasionally show you an actual plan, which you'll get a chance to see as well, and then will even let you edit that plan before the start.

From there, it goes out, it searches the web and it reads the web, searches and reads, searches and reads.

Then it takes what it's actually found by going and doing that search and by reading that content and then finds the next lead.

So it needs to get the basics, And then from there, it can figure out what research is still required to figure out the rest of the answer that you're looking for.

Then once that's done, it finds that answer and brings it all together with sources.

And that's the big thing.

It is all about the sourcing of it, and it's the why I like to use the research tool.

So where you have regular search for a lot of these tools, that's built for quick answers, but deep research is you giving the tool the time.

You're saying, hey, go out, go spend some time kind of getting my answer, but also following leads and then bringing those sources all together.

And there's something that you need to keep in mind when it comes to this, of course, which is that deep research can use more of whatever usage allowance you have.

So some apps will give research.

You'll need to check this out depending on what you have.

Some are going to give research its own limit.

Claude, though, will count against your regular usage.

So you want to kind of, before you go out and do it, or perhaps after you try it out, get a good idea of how much of your usage credits it's actually taking up.

Now, OpenAI actually has some advice for using deep research, as it is called.

The rule of thumb is use search for quick facts and save deep research for depth and thoroughness.

So if you're looking for more than just the basics, that is where deep research can come into play.

So...