What Do I Lose If I Give This to AI?

The question every creator should ask before automating their work.

When I spoke at Open Sauce, I was on a panel and one of the questions that came up was: What do you automate with AI? What’s okay to automate, and what’s not?

I had this whole spiel about how, whenever you’re thinking about automating something with AI, you need to ask yourself one question:

What do I lose if I automate this?

With Future Tools, for instance, I could automate the entire thing. I could have an AI scraper scan the web, look for new AI news, and automatically pull it onto the website. I could have the site completely automated, never touch it, and let it update itself with new news every single day.

But what do I actually lose?

I no longer keep my finger on the pulse. I’m no longer personally reading the news every day. I lose the oversight. I lose the process of staying close to what’s actually happening.

I could do the same thing with YouTube videos. It wouldn’t be hard to create an automation that scrapes the internet every morning, looks for news, writes a script, and hands it to me. I throw the script into the teleprompter, recite it, and the video is made.

But that robs me of the research, the learning, getting into the weeds, and trying to understand what’s going on for myself.

That’s the filter I keep coming back to: What part of me is gone if I give this to AI?

The More I Use AI, the Less I Want It Thinking for Me

Mostly, I use AI for software development. I’ll also toss a video idea into it and say, “Hey, here’s a rough idea for a video, but I haven’t really fleshed it out. What other elements could I add to make it more of a video?”

But, weirdly, I’ve stopped brainstorming with AI as much as I used to.

I’m a little concerned that I’ll get to a point where I stop having original thoughts without AI sort of implanting them in me. The more I use it to generate ideas, shape arguments, and give me language, the harder it becomes to know where the AI ends and my own thoughts begin.

The Hank Green Controversy Wasn’t Really About Using AI

I don’t necessarily think people care that much that Hank Green uses AI. In fact, I don’t even think it’s a secret that he has used it.

He recently built an app around the Artemis mission. There were tens of thousands of photos taken during the mission. They weren’t necessarily labeled by where they were taken along the path, but they did have metadata showing the date and time. Hank built software where you could click anywhere on the mission timeline and see the photos taken around that point.

He made a whole video explaining that he and Claude built it. He used Claude Code to help develop the thing.

There was basically zero backlash. Nobody seemed to care. There wasn’t a huge wave of, “Oh, Hank uses AI? What the heck?”

But then, in another video, he said something like, “I appreciate the pushback,” and people immediately said, “That’s an AI phrase.”

I think that’s what people were actually frustrated by. It wasn’t necessarily that Hank uses AI. It was the fear that AI was starting to seep into his brain.

Hank has always been a science educator. You go to him because he has interesting, well-researched takes on random science topics. He seems well-read. He seems informed. You feel like you’re getting his unique thoughts.

As soon as you start to think an LLM might have written those thoughts for him, the relationship changes. You start wondering, “Am I getting Hank’s thoughts, or am I getting ChatGPT’s thoughts?”

He can say, “The AI didn’t write my script. I only used it for research. These are my own thoughts.”

But even then, people start asking a fair question: If you’re using ChatGPT constantly, aren’t you eventually going to start sounding like ChatGPT?

You might genuinely believe you’re sharing an original thought. But maybe you read that thought in a ChatGPT response a week ago, forgot where it came from, and then repeated it into a camera as though it came from you.

Even without AI-written scripts, heavy AI use can still influence your phrasing, your framing, and the conclusions you reach. So the audience is left wondering how much of the creator they’re actually getting.

Are Any of Our Thoughts Really Original?

Of course, there’s a counterargument here: Are any of our thoughts truly original?

Almost everything we think and say comes from everything we’ve consumed and learned over our lifetimes. We listen to podcasts. We watch YouTube videos. We read books and articles. We have conversations with friends. Now we also talk to ChatGPT.

Is any of it original, or are our thoughts just an amalgamation of all those inputs over time?

AI isn’t completely different from that. But I do think there’s an important difference.

Traditionally, information comes in as a slow trickle from lots of different sources. You might read a book, watch a YouTube video, listen to a podcast, and talk to a friend about the same topic. Because the ideas came from four different directions, you form your own opinion. Your view isn’t the exact opinion of any one person. It’s the culmination of all those inputs.

With ChatGPT or Claude, it’s more like a mega-injection of content and information from a single source.

You ask the question and, boom, this is the answer it gives you. This is the framing. This is the output.

And how was that output shaped? By the average of what was collected across the internet. The model was trained on the extreme view on one end, the extreme view on the other end, and everything in the middle. Its answer will often pull toward the median.

Had you gathered all the pieces yourself, maybe you would have landed closer to one of the extremes. But if AI is constantly synthesizing the world for you, you may keep getting pulled toward the same averaged-out middle.

Then there’s the sycophancy problem.

Most people aren’t going to AI and saying:

“Here’s a thought I have. Give me the strongest counterargument for why this is a bad thought.”

If you’re really serious, you want to hear both sides. But most people use AI as a validation machine:

“Validate this for me. Make me feel like this is a good idea.”

That’s not how I want to use it.

Every Creator Needs to Find Their Own Line

I don’t think there is a blanket take that works for everyone.

Hank’s position was that every creator should have an AI policy and put it in their video descriptions: Here’s what I’m okay using AI for. Here’s what I’m not okay using AI for. If you’re watching my videos, you know what you’re getting.

His own policy included that his scripts would never be written by AI and that none of the images or videos in his content would be AI-generated.

I think every creator needs to find their own line.

It comes back to the same question:

If I use AI here, what do I lose?

Where is the line where you’re willing to give something to automation without taking away from your process or removing what makes you uniquely you?

This matters because I have a strong opinion that YouTube channels offering pure information without commentary or humanity are going to struggle. I’m not necessarily talking about every faceless channel. I’m talking about generic, “Here’s some information for you” channels with no real human element.

People can already get that information from AI.

What they can’t get from AI is you.

Where I Draw the Line

For me, I’m okay using AI to research. I’m okay asking it to show me both sides of something.

If I’m reporting on the news, I want to understand what the proponents are saying and what the opponents are saying. Then I can form my own opinion based on both sides. Sometimes I really disagree with the people who hate something. Sometimes I really disagree with the people who love it. My opinion is probably going to land somewhere on that spectrum because I’ve looked into both sides.

I don’t want to be a talking head for the big companies. I don’t want to be a Google cheerleader, an OpenAI cheerleader, or an Anthropic cheerleader. I want to say:

• Here are the pros.
• Here are the cons.
• Here’s what you need to know.

If I just start saying, “Google launched this. Isn’t it cool? OpenAI launched this. Isn’t it cool?” I’m going to lose people’s respect and trust. They’ll start to see me as part of the PR process, and that’s not what I want my channel to be.

Could I go onto Twitter, read every argument on both sides, find every relevant article, and piece it together myself? Yes. But AI can pull that information together way, way, way faster.

And let’s be honest: How much of the stuff I’m reading on Twitter was probably written by AI in the first place?

I constantly see posts—even anti-AI posts—that I can tell were written by an LLM. So even when you’re scrolling sources that aren’t ChatGPT, AI-generated responses are probably still getting injected into your head.

It’s almost impossible to escape now.

What matters is having a process you trust. For me, that process is using AI to help me see both sides without asking it to hand me the final opinion.

I also don’t like using AI to validate my own ideas.

I don’t want to ask, “I have this idea for a software company. Should I do it?” Nine times out of ten, AI is going to tell me I’m a genius and I should definitely do it. That’s just what AI does, and I don’t think it’s good for that.

I’d rather jump on a phone call with a friend, producer Dave, or someone else in my circle and stress-test the idea with real humans.

What do I lose by asking AI instead?

The nuance of an actual human voice.

So my line looks something like this:

I’m comfortable using AI to:

• Help with software development.
• Research the strongest arguments on both sides.
• Pull information together faster.
• Reorganize a brain dump into a more logical outline.

I don’t want AI to:

• Generate my opinions or takes.
• Write my scripts for me.
• Tell me whether my ideas are brilliant.
• Fully automate the parts of my work that keep me informed and close to the subject.

I have gone to AI, brain-dumped a bunch of ideas, and said, “Outline this into a video.” But in that case, I’m getting zero opinions, zero commentary, and zero takes from the AI. I’m giving it my thoughts and asking it to organize them.

I might record a voice memo and say:

“These are my thoughts. Don’t change my words. Don’t change how I said anything. Just rearrange them so they follow a more logical flow than my random, all-over-the-place rant.”

That is still within my personal acceptable use.

And, honestly, my line has moved.

In the early days, I was willing to give more to AI. The more I’ve learned about AI, how it works, what’s going on under the hood, and the discourse around it, the less I’m willing to give it.

That’s probably not what the AI companies want to hear. But there’s less and less I want to hand over because I want more and more of my own brain in the mix.

The Ethical Objections Are Real

It isn’t only AI output that people are frustrated with. There are also concerns about data centers and where the training data came from.

A big part of the discourse around Hank Green was: “I can’t believe you’re using tools that were trained on stolen information.”

That’s one of the major narratives. AI companies scraped huge portions of the internet, took what they wanted without consent, and trained models on it. Now, when you ask one of those models a question, you’re getting a response from a machine built on top of everybody else’s work.

The other narrative is about data centers. Every time you chat with one of these systems, that request goes to a data center. The more people use them, the more data centers are needed. The more data centers are built, the more communities worry about the water and electricity those facilities consume.

That part is especially relevant to the Hank Green conversation because one of his brands is Eco Nerd. He’s very eco-conscious, aware of climate change, and vocal about preventing it. So some people see his use of LLMs as hypocritical. In their eyes, he’s using systems that rely on data centers that may harm the planet and models built on information taken without consent.

You could argue that a human can read blog posts, watch YouTube videos, and regurgitate the average of what they came across. Why is an LLM different?

I’ll tell you why it’s different: It can do it with the whole thing.

We can only do that with a very finite amount of information. There is a pretty big difference in scale.

So yes, the fear that AI is frying our brains or seeping into our language is part of the conversation. But it isn’t the only part.

Both Sides Need More Grace

Everybody needs to figure out where to draw their own line. But I also think people need to give other people more grace with this stuff.

AI feels less new to people like me because I’ve been playing with it for six years. But to most of the world, this is still very, very new. People are still trying to figure out where it fits into their lives.

Three years ago, software developers were saying AI was crap. Nobody was going to use it for software development. It wasn’t good enough. It would never replace coders.

Fast-forward to 2026. I was at Open Sauce a couple of weeks ago, and every coder on stage was saying, “Man, AI is better than me now.”

The stuff shifts. It’s going to shift again. People who hate it now may change their tune. People who love it now may also change their tune. This is an evolving, ever-moving thing, and everybody is trying to figure out where it fits into the mix for them.

My biggest frustration is the hate on both sides.

On one end, some people find out that someone used AI and immediately try to cancel them:

“You’re ruining the environment. The world is worse because of you. It’s all your fault.”

There’s no responsibility placed on the giant tech companies building this infrastructure. It’s Hank Green, the one person who used AI in a video, who’s treated as the problem.

I think that’s a bad way to look at it. He’s a guy trying to figure this stuff out too:

• Is this ethically okay for me?
• Is it not ethically okay?
• Is it hurting my production?
• Is it improving my production?
• Where does it fit into my process?

But the opposite end of the spectrum is a problem too.

You have super pro-AI people who don’t want to engage in a healthy debate with people saying:

• I don’t like how these models were trained.
• I don’t like that data centers are being built in communities that don’t want them.
• I don’t like that those facilities may consume local water and electricity.
• I don’t like being told this technology is going to take the career I’ve spent years preparing for.

College students have booed keynote speakers talking about AI. Well, yeah. For the last three years, they’ve been told that AI is going to take their jobs. CEOs have publicly talked about UBI, work going away, and students not knowing what to train for because AI may take all of it.

That’s the narrative many students have heard throughout their entire college careers. Of course they’re going to boo when someone gets on stage and says, “Embrace AI.”

The thing they spent four years studying for is the same thing they’re being told will be ripped away from them.

People on both sides need to hear what the other side is actually saying. We need to find the nuanced middle ground where, even if not everyone is completely happy, most people can at least understand the tradeoffs and get on board.

The problem is that the middle ground doesn’t get clicks. It doesn’t get views. That’s why the discourse online looks like everybody is planted at one extreme or the other.

I Don’t Think AI Eliminates Creativity or Entire Jobs

People need to be a little less stressed about where all of this is heading.

Looking a few years into the future is extremely difficult right now. But when I try, I don’t see jobs disappearing wholesale. I don’t see creative roles going away.

I see tasks within those jobs shifting.

Maybe you use an LLM to take some monotonous work off your plate. But your oversight, decision-making, and judgment still matter. Some of the tasks you used to do may be handled by AI. That doesn’t mean your whole job disappears.

The same thing goes for creative work.

If you’re a creative person, you’re still going to be a creative person. You just have more tools at your disposal, and you decide which ones you want to use.

If you want to be the person who draws every piece of art by hand, paints by hand, and creates without any LLM, I think that’s amazing. I also think that work will hold extra value in the future. People will look at something made by hand, without AI, as more valuable than something generated with AI.

At the same time, filmmakers and big productions may use AI for certain elements and still make amazing things. These two approaches can coexist.

Effort Is Going to Matter More

There are photographers who refuse to use Photoshop. They want to do everything in-camera. The exposure, settings, and composition all need to be right when the shutter clicks. What came out of the camera is what you see.

Then there are photographers who bring an image into Lightroom, adjust the colors, mask different areas, and shape your attention around a focal point. The final photo may not look exactly like reality did.

Those are different kinds of photographers using different tools and drawing different lines in the process.

Look at The Odyssey. It could have been shot digitally, but instead they made these massive cameras and shot it on IMAX film. They did it the absolute hard way even though easier options were available.

I think people are going to put extra value on the level of effort that went into something.

So yes, use the tools that make sense for you. Decide where AI belongs in your process. Decide what you’re willing to automate.

But still learn to make the effort sometimes.

And before you give any part of your work to AI, ask:

What do I lose?