Human in the Loop

Fear and Loathing in AI

• By Howard M. Cohen
• 09/21/2026

We’ve talked often in this space about the need for humans to be in the loop with AI, and recently I suggested that the human may actually be the loop in AI. Now, it seems everyone is coming to fear that AI is a loop around humans’ throats.

Has it Changed the Way You Work Yet?
Seldom has software been as widely applicable as AI.

Scientists are using it to unlock answers they couldn’t surface on their own. Much of the early work in AI centered on how proteins fold, and the pioneering medical work con tinues.

The Fortune 500 are in a desperate race to stay ahead by leveraging AI to improve everything from operations to forecasting to strategy and more.

So, how are you using AI as an individual? And how has it changed you? How has it changed the way you work?

Last century, Marshall McLuhan, Neil Postman, and a group of academics taught us that the media we choose to communicate through change us and our relationship with the world. They spoke of the “Media Ecology” creating a communicative environment in which we all lived.

It stands to reason, then, that AI is creating an “AI Ecology” in which we are similarly changed by how we use AI to augment our lives and our work.

So, how has AI changed the way you work?

For myself, it took some time to get past the guilt of using AI, which now strikes me as being very strange. I actually felt guilty having an AI assistant help with my research. It worked so much faster that I felt I must be cheating somehow.

But in perspective, I don’t remember feeling that way when I first started using search engines. Previously, I spent a considerable amount of time in the local library looking things up and finding needed resources. Actually, I very much enjoyed the process and sometimes remember it with great affection. Then I look at the clock!

My research changed with search engines. I spent a considerable amount of time on them, seeking websites that would contain the materials I needed. Then I spent even more time interrogating those websites to dig out the specific facts I needed to support my writing.

Then I met Claude.

Yes, I’ve made no secret of my preferred LLM. Although it often fails, makes mistakes, and loses track of things, I’ve learned strategies to protect myself from that, and I seldom feel like I’ve lost much anymore. And it has eviscerated the time required for research.

Today, I tell Claude what I’m looking for in as much detail as I can muster. It goes off and finds those websites, and it courses through them, extracting exactly the information I require. It summarizes that information for me, providing precise citations and sources so I can properly attribute the information in my work.

I lost my guilt over achieving such time savings, sometimes days of regained time, when I realized there was little difference between the time when I moved from using a manual screwdriver to a power screwdriver. It changed my entire approach to building and fixing things. It was wonderful.

Similarly, I now see Claude as my power screwdriver. It does the same things much faster.

Other Stupid Human Tricks
The other day, I finished working with Claude Cowork on a software development project. It had been quite the slog, moving through a variety of utility software and services to get what we were working on done. I was exhausted.

Then I remembered what I had forgotten to have Claude Cowork do before we were truly finished. As I reopened the interface, I actually felt a slight twang of guilt having to ask poor Claude to do even more than I had already asked. It was as if I were asking my wife to add another course to dinner.

This matched a question that came up at work. I knew I should have the answer. It was a pretty common issue that most people have no problem with. I found myself actually embarrassed to ask Claude to explain it to me. I didn’t want “him” to realize just how stupid I could be.

Finally, a single word took over my entire consciousness: software.

Claude is software. Amazing software, perhaps. Software that can very capably mimic a person's behavior, but my no means a person itself. In other words, it has no dog in this hunt. It is not personally invested in our relationship. It does not get tired or frustrated. It just does what I ask.

I thought of many people for whom I wished I could say the same.

The Tool is Only as Good as the Craftsman Who Wields It
This is one of my favorite ancient proverbs, and like most, it is anonymous. Nobody can pin down who first said it, though many have paraphrased it.

In less than two years, we've moved from “vibe coding” to prompt engineering, context engineering, harness engineering, agentic engineering, loop engineering, and now graph engineering.

If you asked experts like Boris Cherny and Andrej Karpathy, you’d more likely hear them talk about conversation. Today, we converse with our LLMs. We pose a challenge. They ask questions. We answer. They probe deeper. At some point, they begin work, and we are either thrilled or dissatisfied. So, we iterate. We take the conversation even further. And we keep iterating until we obtain the results we sought in the first place.

This suggests that how well you use AI, and how much value you get in return for your efforts, is dependent upon how well, how precisely, how thoroughly we can describe what we want. And what we don’t want.

Will AI Extinguish the Human Race
With recent spate of social media posts from departing and remaining employees at frontier AI labs, our fear that this technology will end the human race has resurfaced. It is amazing to note how easy it is these days for people to get their 15 minutes of fame with a properly placed “tweet.”

I personally prefer to believe in people. I believe people are adaptable and circumspect enough to recognize what they need to do and when. There are, of course, exceptions. Those who cannot recognize those things usually end up in elected office, and that’s a problem.

Enough brilliant people in the vanguard of AI development have the morals and integrity to say what is true and act on it as necessary. I believe they can and will overcome the manipulators, malignant narcissists, and other government officials and billionaires who threaten to allow AI to ravage the world. Until then, I’m making incredibly good use of the time AI has given back to me.