The Battle of the Claudes

Recently, I was working through a particularly important piece of strategic analysis. Like many people, I leaned heavily on AI to help process the information, identify themes, and formulate recommendations. The resulting document looked great. The writing was polished. The arguments were logical. The conclusions seemed reasonable.

There was only one problem.

Some of the conclusions weren’t actually mine.

More specifically, some of the hypotheses being presented weren’t fully supported by what I saw in the data or what I believed to be true. Had I spent more time sitting with the information and thinking through the implications myself, I would have challenged several of the assumptions. Instead, I moved too quickly.

Looking back, I think the quality of the writing obscured the quality of the thinking. The document sounded intelligent. It looked professional. It was easy to accept because it was presented so convincingly. What I failed to do was spend enough time asking myself a simple question:

Do I actually believe this?

That experience has been rattling around in my head ever since because I don’t think it’s an isolated incident. In fact, I think it may be one of the more interesting side effects of AI that we aren’t talking about enough.

A few weeks later, I was discussing the growing volume of information flowing through our organizations with a colleague. Strategy documents. Analysis. Recommendations. Feedback. Position papers. Emails. Comments. Rebuttals. It feels like every week there’s more of it, and the pace at which it’s arriving continues to accelerate.

During that conversation, I made an observation that immediately resonated with him.

“We’re in the middle of the Battle of the Claudes.”

He laughed because he knew exactly what I meant.

Someone uses Claude to generate a detailed document. I use Claude to summarize it. I use Claude to help formulate a response. They use Claude to critique my response and generate a counterargument. I use Claude to summarize their counterargument and identify key themes. Somewhere in the middle of all of this, two humans review the output, make a few edits, and call it collaboration.

The funny thing is that I’m only half joking.

Before anyone mistakes where this article is headed, let me be clear. I wouldn’t trade these tools for anything. The amount of information I can process, the speed at which I can explore ideas, and the quality of feedback I can get from AI are extraordinary. In many ways, these tools have fundamentally changed how I work for the better. The challenge isn’t that AI is bad. The challenge is that the temptation to delegate judgment increases as the tools become more capable.

The issue isn’t AI.

The issue is that AI has dramatically reduced the cost of creating information while doing very little to reduce the effort required to evaluate, challenge, and ultimately own that information.

Historically, creating a thoughtful piece of analysis required real effort. You had to gather information, organize your thoughts, build an argument, and communicate it clearly. That effort naturally limited how much content people produced.

Today, much of that friction has disappeared.

A person can generate ten pages of analysis in minutes. They can ask for alternative perspectives. They can request counterarguments. They can expand sections. They can create executive summaries. They can continue refining the output until it looks polished, comprehensive, and authoritative.

Unfortunately, the receiving side of the equation hasn’t changed.

No matter how many GPUs NVIDIA ships, I still have twenty-four hours in a day. I still have meetings. I still have customers. I still have projects. I still have decisions to make. Most importantly, I still have a finite amount of mental energy available to process information.

Human attention remains stubbornly human.

That’s the bottleneck.

What’s fascinating is how quickly many of us have adapted to this new reality. Faced with a flood of AI-generated content, we increasingly rely on AI to help us consume it. Someone sends a lengthy document, so we ask AI to summarize it. We receive a detailed proposal, so we ask AI to identify the key themes. We need to provide feedback, so we ask AI to draft a response.

At first glance, this seems perfectly reasonable. In many situations, it is. AI saves time. AI helps us process information more efficiently. AI can help us focus on what matters.

The problem occurs when we stop using AI to accelerate our thinking and start using it to replace our thinking.

That’s where my strategic analysis example becomes relevant.

The issue wasn’t that the AI-generated recommendations were irrational. The issue wasn’t that the writing was poor. The issue wasn’t even that the analysis was objectively wrong.

The issue was that I hadn’t fully formed my own opinion before accepting the conclusions.

Somewhere along the way, I delegated judgment before I had earned the right to delegate it.

That’s a subtle distinction, but I think it’s an important one.

One of the things I’ve noticed over the past year is that many of us use AI to expand ideas more often than we use it to deepen them. A simple thought becomes a three-page document. A straightforward recommendation becomes a detailed dissertation. A brief email becomes an executive briefing package.

The problem is that length and insight are not the same thing.

In many cases, what we’re generating isn’t deeper thinking. It’s simply more words. The core idea becomes buried beneath repetition, elaboration, and unnecessary detail. What could have been explained in a few concise paragraphs becomes an exercise in extracting the actual point from the surrounding text.

Anyone who knows me will tell you that brevity has never been my greatest strength. My brain likes details. I appreciate context. I enjoy nuance. Even with that admission, I find myself reading AI-generated content and wondering whether half of it needed to exist at all.

What’s particularly interesting is that I don’t think most people are doing this intentionally. In fact, I think they’re experiencing the exact same pressure that I am. They’re overwhelmed by incoming information. They’re trying to keep up. They’re using AI to help them process material. They’re generating more content because content has become easier to generate.

The recipient then uses AI to process the content.

Everyone is acting rationally.

The system as a whole becomes less efficient.

I’ve observed this behavior across every level of an organization. I’ve seen it from executives. I’ve seen it from individual contributors. I’ve seen it from Product, Engineering, Marketing, Sales, and Operations. I’ve seen it inside my own company and in conversations with people outside of it.

Occasionally, I’ll read something and think to myself, “There’s no way this is how that person actually thinks.”

Not because they aren’t intelligent. Usually it’s because I know them well enough to recognize that the writing doesn’t sound like them. The level of detail doesn’t match how they communicate. The document feels more like a generated artifact than a reflection of their actual thought process.

I’ve also discovered that AI doesn’t always save as much time as people assume. There have been numerous occasions where I’ve asked Claude to create an analysis, presentation, or recommendation only to spend the next hour correcting assumptions, restructuring arguments, and pulling the output back toward what I actually believed.

In many cases, it would have been faster for me to create the initial structure myself and then use AI to improve the language, visuals, and formatting.

I’ve found AI to be remarkably effective at identifying signals, surfacing patterns, improving language, and challenging ideas.

What it cannot do is replace my context, my judgment, and my conviction that come from deeply understanding a problem.

Those things are earned through experience, customer conversations, organizational knowledge, and spending time wrestling with difficult decisions. AI can inform that process, but it cannot own it.

That’s why I’ve started changing how I approach important work.

For critical analyses, strategic decisions, and recommendations that genuinely matter, I’ve become much more intentional about creating space to think. I still use AI. I still rely on it heavily. But I force myself to spend time with the material. I read it. I challenge it. I compare it against my own experience. I ask myself whether the conclusions actually reflect what I believe.

I’ve also found myself scheduling more conversations. Not meetings. Conversations.

Sometimes the best response to a lengthy document isn’t another document. It’s a fifteen-minute discussion. Five minutes of human dialogue can often resolve what would otherwise become multiple rounds of AI-generated analysis, rebuttals, summaries, and counterarguments.

It’s surprisingly effective.

The truth is, I don’t have a perfect answer. In fact, I suspect this problem gets worse before it gets better. The cost of creating information has collapsed. The amount of information being generated will continue to grow. The tools will continue to improve. The pressure to move quickly isn’t going away.

What concerns me is that we may accidentally optimize ourselves out of the very thing that creates value in the first place.

Thinking.

Not generating words. Not generating documents. Not generating responses.

Thinking.

I don’t believe the answer is abandoning AI. That would be foolish. AI has become one of the most valuable tools in my professional toolkit. But somewhere along the way, many of us stopped using AI to accelerate our thinking and started using it to replace it.

The challenge isn’t winning the Battle of the Claudes.

The challenge is maintaining ownership of our thinking.

Wishing you all the best

Mike