A few months ago, I decided it was time to embrace the future.
Everywhere I looked, people were announcing that AI had changed their lives. Productivity gurus were working only 17 minutes a day. LinkedIn was full of smiling people claiming they had “automated their entire business.” Teenagers were building apps from their phones. Retirees were generating travel itineraries, family histories, and suspiciously professional-looking newsletters.
How hard could it be? I already knew how to use email, Excel, and a rice cooker. Surely I could master a chatbot? I was wrong.
The Oat Milk Incident
One morning, I opened ChatGPT to ask a simple question:
“What’s the best way to froth oat milk?”
This seemed straightforward.
Instead, I discovered that I apparently now had Projects, GPTs, Memory, Temporary Chats, Custom Instructions, Canvas, Deep Research, and enough settings to configure a small government.
By the time I found the text box, I had accidentally created what seemed like a lifelong relationship with an artificial intelligence that would remember my preferences indefinitely. All because I wanted a better cup of chai.
Somewhere in a data center, a server now knows I prefer oat milk, drink chai without sugar, enjoy cryptic crosswords, and once spent forty minutes researching the history of veganism after getting distracted by a completely unrelated clue.
I came for tea advice. I left with a personalized digital companion.
Prompt Engineering: The Art of Talking to Robots
In the early days of computing, we learned commands. In the AI era, we’ve learned diplomacy.
Apparently, you can’t simply ask for something. You must provide context. And examples. And constraints. And objectives. And audience demographics. And desired tone. And edge cases.
I recently spent ten minutes crafting a carefully worded prompt that began:
“Act as a patient technical support person helping a retired engineer understand why there are now seventeen AI models, twenty subscription tiers, and forty-seven YouTube experts claiming the others are obsolete.”
Meanwhile, when speaking to actual humans, I still communicate primarily through incomplete sentences and vague hand gestures. Yet somehow I’ve learned to write 200-word briefing documents for a chatbot to ask why my oven door won’t open.
I’ve become startlingly polite to machines. I say “please.” I say “thank you.” Not because it helps. Because if the robots take over one day, I want my file marked cooperative.
The Productivity Olympics
The internet is full of people demonstrating revolutionary AI workflows.
Every video follows the same plot.
Step 1: Connect Tool A to Tool B.
Step 2: Connect Tool B to Tool C.
Step 3: Declare that you’ve increased productivity by 10,000%.
What they don’t show is Step 2.5:
Spend three hours wondering why Tool B refuses to talk to Tool C because a checkbox hidden inside a submenu wasn’t enabled. The final workflow saves four minutes per week. The setup takes an entire Saturday.
Still, everyone ends the demo looking delighted. I suspect caffeine is involved.
The Confidence Problem
AI has an unusual personality trait. When it knows the answer, it sounds confident. When it doesn’t know the answer, it sounds the same. This creates a fascinating relationship. One minute, it’s helping you summarize a 50-page document in seconds. The next minute, it’s inventing a historical event that never happened while maintaining the cheerful certainty of a tour guide explaining local attractions.
You learn quickly. You trust, but verify. Then verify the verification. Then ask a second AI to check the first AI. Then Google it anyway.
My Career as a Professional AI Student
The first surprise was discovering that there isn’t an AI tool. There are approximately 4,732 of them, and every morning a new one appears.
“Have you tried this one?” someone asks.
“No.”
“It’s amazing.”
“What does it do?”
“It does what the other one does, except faster, with agents.”
Naturally, I nod as if I know what agents are. I do not know what agents are. I have watched fourteen videos explaining agents. I am now less certain than when I started.
The biggest surprise wasn’t learning the tools. It was learning that learning the tools never ends. Every week there’s a new model. Every month there’s a new feature. Every day, someone posts:
“Everything has changed.”
Again.
I have lived through punch cards, floppy disks, Windows updates, smartphones, cloud computing, social media, streaming services, and whatever TikTok is doing. Yet I’ve never seen a technology move quite this fast.
Just when I think I’ve caught up, another announcement arrives explaining that the model I learned yesterday has been replaced by a newer one that reasons better, thinks longer, works with agents, and can apparently generate a podcast discussing the results.
At this point, AI feels less like a tool and more like a treadmill set slightly faster than my walking speed. I’m making progress, but only because stopping would send me flying backward.
What AI Is Actually Good For
After all the experimentation, I’ve reached a surprisingly simple conclusion.
The best use of AI isn’t replacing human thinking.
It’s replacing human drudgery:
- First drafts.
- Formatting.
- Summaries.
- Brainstorming.
- Research assistance.
It is the digital equivalent of chopping onions before cooking dinner. It handles the tedious parts so you can focus on the interesting parts.
Where AI has genuinely impressed me is in helping me research purchases that would otherwise require hours of reading reviews, comparison charts, and conflicting opinions from strangers on the internet. Recently, I used it to help narrow down the best gate-access intercom system for our property and the best Wi-Fi extender to reach those stubborn corners of the house where the signal goes to die. Instead of spending an entire weekend opening forty browser tabs, I spent an hour asking questions, comparing options, and learning what actually mattered. AI didn’t make the final decision for me, but it did a remarkably good job of helping me ask better questions.
The danger begins when you ask it to make decisions that require judgment, experience, common sense, or responsibility. In other words, the things we were trying to avoid doing in the first place.
AI can tell you the pros and cons of ten intercom systems. It cannot tell you whether your spouse will hate the one that chimes like a foghorn every time a delivery driver arrives. It can compare Wi-Fi extenders. It cannot explain why your router works perfectly until guests come over and everyone starts streaming videos at once.
The real skill isn’t letting AI do your thinking. It’s letting it do the homework while you remain responsible for the answers.
Final Thoughts From the Learning Plateau
Are AI tools worth learning? Absolutely. They’re useful, impressive, and occasionally astonishing.
Will the experience feel like assembling IKEA furniture while the instruction manual updates itself every fifteen minutes? Also yes.
Some days you’ll feel like a genius. Other days, you’ll spend an hour trying to remember which chatbot generated which document in which project inside which workspace. And that’s before you start comparing subscription plans.
The future has arrived. It’s exciting, transformative, occasionally bewildering, and constantly asking whether you’d like to upgrade.
As for me, I’m still learning. Right after I figure out how long to froth that oat milk.
Leave a Reply