I’m more productive than I ever have been in my life, but I’m not happy. I used to enjoy starting a new project, but now it all feels like, “What’s the point? AI can just do it.” It’s time to start adapting! This is my story.
As with the majority of the tech world, I’ve been using AI to automate all the things. For the past three years or so, I’ve gone from this:
To this:
I now have a CLI for just about any task I need to do online, both personal and professional, and a framework to simply ask my AI agent, “Hey, build me this CLI tool to work with X.” It has the necessary knowledge to do the research, follow my workflows, authenticate for me (since everyone needs a lastpass CLI so their AI agents know all of their passwords), and save the work so I never have to authenticate again.
I have my Hermes Agent running on my Mac mini, managing my incoming channels like email, Slack, and iMessage so I can give my agents instructions from any medium.
I have skills for any specific knowledge set my agents need, scripts to accompany those skills and help my AI agents, subagents for when my main agents need some help, and workflows defined for just about everything.
But instead of feeling good about my work, I feel more like a middle manager herding cats, directing my AI agents without thought and with just a broad idea of what I need. Why sweat the details when your trusty AI agent can handle it? Mundane. Or so I thought.
Why even open an IDE or a browser when you can simply ask in natural language? When your only interface is a natural language prompt, the world feels less fulfilling.
If AI can do it, what’s the point of trying?
As an automator, you know, Adam, “the Automator,” automation has been in my blood for a long time and long before AI. When I was working in IT, I’d spend hours automating a process that’d take me a few minutes, mostly all in the name of efficiency (or the thought of it). I enjoyed the process of logically breaking down a problem, making incremental changes to a script, and seeing it deployed into production. It felt rewarding.
I did that. I was the one that figured that out. I built that. Now? That rewarding feeling that used to rush into my brain after those long debugging sessions is gone.
Now that I’ve built my personal “agent OS,” I can one-shot a request and be done with it. Great! Easy! But fulfilling? Nope. Being able to push a button and get back a result better than, or nearly as good as, what you can do is efficient but will never trigger that dopamine release every engineer needs to be happy.
“It’s supposed to be hard. If it wasn’t hard, everyone would do it.
The hard is what makes it great.”
-Tom Hanks (as Jimmy Dugan) in A League of Their Own
I applaud anyone non-technical getting into vibe-coding, building software from prompting rather than writing code. I believe more ideas will turn into real software now that people can build from prompts instead of writing every line of code. Hell, I’ve spent over two years myself, pretty much vibe-coding an app called BrickBuddy, which I use in my LEGO reselling business, that I can’t live without now.
But if anyone in the world with $20/month can build the same apps I do, the excitement isn’t there. The euphoria an engineer gets when releasing an app to the Apple App Store or getting their first dollar of revenue from a SaaS app doesn’t feel the same to someone who truly cares about their practice.
That’s where the temptation starts. If AI makes building feel easy and cheap, it’s hard not to wonder whether it can turn that speed into money.
Let’s get rich…with AI!
The AI bros love how easy it is to create with AI. They pump out content, apps, courses, and anything that can be automated as fast as possible, and love spamming quantity in search of the next dollar. They don’t care about what it takes to create the product. They don’t care about the journey, only the destination.
I can see this to an extent. During OpenAI’s 2x promotional period, where I had tokens to burn, I decided I was going to get rich through affiliate marketing and was going to let my AI agents build the entire system with no manual intervention by me. After nearly 30 hours of continuous work, I had nothing to show for it. It was a failed experiment because I didn’t set up the guardrails, set the rules, or establish a framework.
I’m sad to say that “Make me rich. No mistakes. ASAP” is not a great prompt.
The AI automation addiction
Now, my automation has turned into an addiction and not in a good way. Every time I begin a task on my trusty laptop, my brain automatically goes to AI.
“Should I just create a skill for this task and have Codex do it?”
“Why am I doing this by hand when I can just ask Claude Code to handle it? It has access to everything I do.”
“This would be so much faster if I simply offloaded this onto Hermes”
And I probably give into these urges 99% of the time, judging from my nearly 100 CLI tools that I’ve built and over 500 skills.
I’ve built up quite a knowledge base where it doesn’t matter what the agent harness is; Codex, Claude Code, Antigravity, Hermes, OpenClaw…my tooling is universal. Why? Because after one agent harness usage runs out, I can easily transition to the next without skipping a beat.
You may think, “that’s great, Adam! You’ve been able to automate all of the mundane stuff in your life!” True. But it’s only after you automate all of the tasks you think are mundane, and you don’t have any tasks left, that you begin to think, “Is my entire career just mundane enough to be completely automated?” What do I actually bring to the table?
As I’ve been writing this exact article about the drawbacks of AI, my brain has gone to AI editing and having it finish it. I think, “I have enough of my voice in this already. I bet it could finish it and clean this up much better than I could. Why am I still writing?” It’s the same thoughts I had when I was writing PowerShell scripts every day, but applied to everyday life!
How AI Transformed a $40K/month Blog to <$1,000
Remember Stack Overflow? That site where millions of people went to for help or a snarky response from Internet strangers? Check out this graph.
Doesn’t look too good, does it? Now, look at the adamtheautomator.com blog’s traffic graph from 2020 to now.
Looks terrible. From a peak of 1.2M views/month to around 70K; a drop over 90% over a few years. Granted, this behavior isn’t entirely dependent on AI, but when’s the last time you typed into Google ‘how to delete a file with PowerShell’ vs. simply telling your IDE to “delete this file now”? What’s easier? Firing up a browser, typing in the query, sifting through web page results to eventually find the answer and it works half the time, or telling your AI agent to do it once and it’s done? Yeah, I thought so.
Tutorial-based blogs like this one have seen a dramatic fall, and rightfully so. I don’t read blog posts anymore. Hell, I don’t even read webpages anymore. My information is purely through the filtered, summarized output of AI. Who’s got time to sift through all of that unnecessary cruft anyway?
I firmly believe in market innovation. If I’m on the wrong side of that innovation, it’s time to suck it up and innovate with it. Yeah, it sucks, but I’m not out here protesting AI data centers because it’s taking our jobs. AI is here to stay, and we must adapt or die.
Turning Things Around (Hopefully)
I love AI agents. Many years ago before ChatGPT, I remember saying that AI feels like automation 2.0. It’s taking the game up another abstraction layer and allowing people like me to automate tasks never possible before with deterministic code.
I still love automation and AI, but I have to get out of the habit of relying too heavily on one-shotting my tasks. My natural tendency is to automate everything. This tendency once worked to keep me feeling fulfilled, but not anymore. I need to understand when Adam’s brain needs to be involved and stop at that point.
I’ve gotten great at building tools and instructions for AI agents and need to continue this. Because of my decades of experience in automation and workflow-building, I understand workflows, frameworks, and architectures that work. I need to continue building upon this and infusing this knowledge into AI skills.
I have a love/hate relationship with AI. It destroyed my biggest revenue stream and robbed my precious dopamine hits, but it has opened the door for the next chapter of my career.
I’ve already established some consulting projects around AI automation and have enjoyed them. My work hasn’t changed much, but the way I work has, and it’s time I started adapting and got out of my insanity rut of doing the same thing over and over and expecting a different result!
Made by Adam ©, in Indiana. Edited by AI.
