This isn't another AI hot take babe, just a personal take.
I need to begin with a disclaimer because apparently talking about AI on the internet now requires the same protective equipment as handling radioactive material. This is not the correct opinion on AI. I do not possess the sacred glowing USB stick containing the future of humanity, and I am not here to tell you AI will save civilization, destroy civilization, steal your lunch, cure every disease, become your manager, or invent another subscription service we all forget to cancel.
This is just my personal take after spending the last few years watching AI tools go from "haha look what this weird chatbox wrote" to "okay apparently this is part of my workflow now." There has been a little chaos, a little calm, and a concerning number of browser tabs.
¯\_(ツ)_/¯
Mostly, I think AI is a tool. A ridiculously powerful, occasionally brilliant, occasionally confidently-wrong tool. And like most tools, I don't think the interesting question is simply "Is it good or bad?" The better question is probably: What are you using it for, and did you remember to keep your brain plugged in while you were doing it?
Son, let me tell you of the before times
puffs a pipe that randomly appears from nowhere
There was a time when I was learning coding in university and 2AM meant staring directly into the void because my code did not work. Not kind of didn't work. It did not work, the error messages were multiplying, and the computer showed absolutely no concern for my emotional wellbeing.
Our AI assistant at the time was something called pasting the error directly into Google and hoping another programmer had experienced exactly the same catastrophe seven years earlier. You would copy the error, paste + enter, and send your little distress beacon into the night, hoping some Stack Overflow veteran would spot it flickering on the horizon and dispatch the warriors of the comment section.
People helped people. Sometimes you would stumble across a thread where someone had already had the exact same problem, which honestly felt like discovering a handwritten survival note left by a previous traveller. Then one terrifying keyboard shaman with 146,000 reputation would appear with a six-line explanation that made you realize you had misunderstood something fundamental for approximately three weeks. And before that, the IDE elder veterans didn't even have what we had. They had books. Manuals. Printed documentation. Physical objects they had to open with their hands—and presumably hair to tear out.
Respect.
Now, sometimes you can paste an error into an AI tool, press enter, and get a completely reasonable solution almost immediately. That is wild. Past-me would have considered this forbidden magic. Current-me treats it more like an extremely fast junior dev who has absorbed an alarming amount of documentation, never sleeps, and occasionally looks me directly in the eyes before making something up. Useful, but supervision is recommended.
Let me tell you about the phases
I have realized my relationship with AI has happened in phases. These are not official phases. There was no committee. Nobody laminated a roadmap. This is simply the taxonomy my brain has created after several years of observing the circus from somewhere inside the tent.
Phase 1: Grand Entrance
COVID is wrapping up and AI makes its grand entrance into the corporate landscape. The doors swing open, the metaphorical fog machines activate, and somewhere in management a voice rings out:
IMPLEMENT! RIGHT AWAY!
Meanwhile, those of us actually trying to use the thing are sitting there thinking, implement what, exactly? The quality was genuinely rough in places. The IDE is freezing. The chatbot is unresponsive. Our laptop RAM is crapping the bed. The back of the laptop is so hot the desk is beginning to warp into a new geological formation, and we are apparently supposed to be revolutionizing productivity while wondering whether the machine itself is about to achieve nuclear fusion.
My personal favorite is when the model decides, "Hmm, this looks like something you don't need in your .gitignore. Let me do you a solid bud and remove it."
RIP.
There is something spiritually humbling about adopting a shiny futuristic technology and then spending the afternoon recovering from the shiny futuristic technology deciding that your protective configuration file looked optional. The enthusiasm was enormous, but the actual experience could be clunky, unreliable, or just plain weird. Attempts were made. Demos were shown. Experiments occurred. Then a bunch of those experiments quietly migrated into folders nobody opened again.
Phase 2: Managers Persist
But management did not forget.
Oh no.
They remembered the chatbox.
So we turned our attention to it once more. This time we struggled more deliberately and with gusto! We prompted, re-prompted, adjusted instructions, argued with outputs, explained what was wrong, and occasionally found ourselves teaching the chatbot why its own previous answer had gone sideways. At certain points it felt less like having an intelligent assistant and more like coaching a very enthusiastic junior dev who had consumed the entire internet overnight and was now absolutely bursting to show you what it learned. But then something interesting happened: we found things that actually worked.
Tedious little tasks that used to swallow chunks of the day could suddenly move much faster. Boilerplate, summaries, documentation starting points, explaining unfamiliar code, transforming data, brainstorming edge cases, organizing messy notes, or simply getting the first ugly draft of something onto the screen all became viable uses. The chatbox stopped being just a novelty. There were genuinely useful tools hiding in there if you were willing to spend enough time figuring out which jobs they were good at and which jobs caused them to wander into the woods.
Phase 3: The Savvy Divide
Then the divide appeared. A handful of people continued wrestling with the tools. They tried things, broke things, tested different models, learned how much context mattered, got better at prompting, and slowly figured out where AI was legitimately helpful versus where it produced flaming garbage wearing a tiny business-casual blazer. The other half crossed their arms and turned up their noses.
And listen, I understood some of the hesitation. I still do. There are completely fair reasons to question the technology, especially once you get into privacy, ownership, data, jobs, energy use, creative work and the deeply scientific technical category known as what the heck is this thing doing now?
But I decided to keep my head down and learn what I was being told to learn because, personally, having food is nice. Big supporter of food. Huge fan. Also housing. And taking care of pets. So while everybody else was debating whether AI deserved to exist, I was somewhere nearby clicking buttons like, "Okay weird robot. What does THIS button do?"
Phase 4: Rampant Hot Takes Everywhere
Then came the hot takes.
Ohhhh, the hot takes.
Quick, close LinkedIn before it sees you.
Suddenly the feed felt like two competing street preachers had been handed unlimited Wi-Fi. On one side, AI was going to replace every human profession by Thursday afternoon. On the other, AI was useless, morally bankrupt, technically unimpressive and apparently responsible for the collapse of Western thought.
Guys. Please. I just wanted to learn how the tooling worked.
The conversation became weirdly personal. Complete strangers were going to war in comment sections because one person used Copilot and another person had a drawing tablet. Clicking the little x on that tab could not come fast enough.
And while all of that was happening, the discussion stopped feeling theoretical. Tech companies were cutting huge numbers of jobs. People were scared. I was scared, and I am still kinda scared. Sometimes it feels like if you aren't Gilfoyle from Silicon Valley, sitting in a dark room casually rebuilding an entire infrastructure stack before lunch, you are screwed.
The public conversation tends to spotlight the extremes: the people sprinting toward AI screaming that everything is about to change forever, and the people sprinting in the opposite direction screaming that everything is about to change forever. We don't hear nearly as much about the rest of us swimmers just trying to stay afloat in increasingly sus water.
Then there was the everliving price hike situation, which had some early-adopting companies absolutely shooketh. Everyone had rushed to bolt these shiny new tools into workflows, only to discover that apparently unlimited magical robot assistance had invoices attached. Suddenly the same organizations yelling IMPLEMENT! were staring at usage charts like somebody had left the office thermostat set to 31 degrees all winter.
And the AI conversation escaped the neat little fenced area of software development anyway. Code wasn't enough. The monster looked around at writing, images, music, video, education and creative work and apparently said, "Oh good, a buffet."
Phase 5: The Resistance
Can I quickly add that throughout all of this I am still just... here. Learning. Trying to keep up. Opening documentation, closing documentation, opening it again because apparently I did not absorb it through osmosis. By this phase, we are back to reading documents like litigation professionals. Thousands of lines of markdown before 9AM and we haven't even touched a bracket key yet. There are changelogs, model cards, API docs, tool descriptions, security notes, deprecations and seventeen slightly different ways to accomplish something that did not exist six months ago. There are mornings where I feel less like a developer and more like I am preparing exhibits for a trial involving the State versus JavaScript.
I personally like some parts of my life far away from computers. Music is one of them. There are moments when I do not want an algorithm optimizing anything. I want somebody playing an instrument slightly imperfectly because they are a person with wrists and feelings. But I still learn about the technology because, returning once again to my previous economic thesis, eating is good. So is taking care of pets.
At the same time, resistance became much more visible. Communities pushed back against data centres. Artists pushed back against AI-generated work. People questioned training data, attribution, ownership, jobs and whether every technically possible thing needs to immediately become a product. And somewhere inside this strange cultural soup, the World YouTube channel dropped this video baddie, If You're Human And You Know It (Show It). It's a jam. The whole situation captures something I think has shifted in the conversation: we are no longer only asking "What can AI make?" We are also asking "What do we still want humans to make?" Those are not the same question.
Phase 6: A Line in the Sand... Maybe?
Which brings me to where I currently sit. I think we as humans have to decide which parts of AI we actually want in our lives. Not which parts exist. Not which parts a company wants us to adopt. Not which parts somebody on LinkedIn promises will make us 7000% more productive while generating passive income from a yacht. Which parts we want. Where is the line for you? Mine is still moving.
I like AI as a tool, but I am also sooo freaking glad I learned in school without it. Skill atrophy is real, or at least it certainly feels real when I notice myself reaching for assistance on something I know perfectly well I used to work through manually. That bothers me enough that sometimes I deliberately go backwards.
For one of my projects, a quirky free online multiplayer game called Oddly Specific, I went back to a time before the computer. I drew the game logic out on cookie sheet paper. Cookie paper? Parchment paper, I think. You know the stuff. I will post it once I find it.
The point was that the game needed to make sense on that paper, and it needed to make sense inside my own brain, before a line was written in React. There is something strangely useful about forcing software logic to survive contact with a pen, a giant sheet of kitchen paper and no autocomplete.
I support getting assistance from robots for repetitive daily tasks, and I hope that one day they don't come for me.
Fun fact: the first rendition attempt of Octivary made AI "crap itself in italics." I tried it for fun. It wasn't ready, so I put the chatbox down.
So I slogged through math. Actual math. PTSD and JSON files. And eventually got the first POC alive and running.
There was something deeply satisfying about that. Not because doing everything the hard way is morally superior—I absolutely do not believe that—but because there are some moments where wrestling with the problem yourself teaches you where its bones are.
Okay, but isn't using AI cheating?
This is one of those questions people new to AI ask where the internet tends to offer answers with all the subtlety of a frying pan. My answer is: depends what you agreed the work was supposed to measure.
If I'm taking a programming course where the assignment exists specifically to test whether I understand recursion, asking AI to solve the entire assignment defeats the point. Congratulations, the robot understands recursion now. I do not.
If I'm at work and need to convert 40 repetitive chunks of text into a predictable format and AI can safely shave an hour off the task? Baby, automate the boring thing. The difference is context. AI is a little like bringing a calculator. A calculator in an accounting office? Wonderful. A calculator during an exercise specifically designed to see whether you understand the calculation yourself? Different situation. The tool does not become magically ethical or unethical by existing. The rules, expectations, data involved and purpose all matter.
How do I know when the AI is wrong if I'm asking because I don't know?
THIS. This is one of the least satisfying problems in the entire AI conversation because if I already knew the answer, I probably would not be asking the machine. AI giving me ten funny names for a fictional raccoon detective? Low stakes. Release the raccoons. AI giving me something involving money, security, health, legal issues, production infrastructure or a major technical decision? Now we verify. Check primary documentation. Run the code. Write tests. Compare outputs. Get another human involved when necessary. ELI5: treat AI like your friend who says, "I know a shortcut." Sometimes your friend knows an excellent shortcut. Sometimes you end up standing behind a Costco beside a locked fence wondering how this became your afternoon. Verification matters.
Will using AI make me worse at the skill I'm using it for?
Possibly. There. Horrible answer. But I think it depends enormously on which part of the thinking you outsource. If AI handles repetitive formatting after I've already made the important decisions, wonderful. If AI helps me understand unfamiliar code but I still trace through what it is doing, also useful. If AI produces every implementation, every test, every explanation and every decision while I sit nearby clicking Accept like a pigeon operating a vending machine, I am probably not learning very much. And I've felt this personally, which is why I deliberately keep some things manual.
Test-driven development has saved my butt more than once
I was lucky enough to be trained under the philosophy of test-driven development, so test writing is deeply embedded in how I think about building software. I still like writing important tests myself without assistance because those tests catch A LOT. Capital letters intentional.
I still believe heavily in regression testing, automation, linting, edge-case scripting and generally putting enough guardrails around a system that when something explodes, it explodes somewhere controlled instead of at 4:47 PM on a Friday.
Early in AI-assisted development, I absolutely got burned. Young Padawan Me was basically:
"OH THIS CHANGES THE GAME. HERE WE GO. DO MY WORK. WAHOOO!"
Narrator: it did not, in fact, simply do all the work. The code looked good. The explanation sounded good. Sometimes everything even ran.
Then the tests walked into the room carrying a clipboard. Absolutely not. Wrong assumptions. Missing edge cases. Subtle regressions. Libraries used strangely. Functions that technically handled the example while quietly failing the actual requirement.
Classic: "It works on my machine." Fantastic. Unfortunately, we do not distribute your machine to the entire customer base.
Those failures taught me something useful: don't throw the human out of the loop. Whatever you do, somebody needs to remain present who can look at an output and say, "Hey... that's weird." That sentence might actually be one of the more important technical skills of the AI era: recognizing weird, knowing enough about the system to feel when something does not quite fit, and being willing to investigate instead of accepting the beautifully formatted answer because it arrived quickly and used confident punctuation.
Oh no. This is becoming a hot take.
Ew. Abort.
I did not come here to deliver commandments from the AI mountain. Really, I'm just here to say that if you want to play with fire, there are different kinds of fire, and it helps to know which one you're lighting.
You can use AI to explain, summarize, prototype, research with verification, brainstorm, automate repetitive work, generate starting points, translate between formats, help navigate documentation, challenge an idea, find possible edge cases, or rubber-duck a problem when every human you know is asleep.
And you can also decide not to use it. That is an option too.
Hopefully the Choice Compass can help narrow down which tool—or whether any tool at all—makes sense for what you're trying to accomplish. I don't think everybody needs to become an AI power user, and I definitely do not think every creative act needs optimization.
Not every inconvenient five-minute task requires seventeen agents, four MCP servers, a vector database and an automation named FINAL_v2_REAL_FINAL_USE_THIS_ONE. Sometimes you can just do the thing. The point, to me, is choice. Intentional choice.
TL;DR
AI tools are neither magical digital baddies sent to rescue civilization nor one giant evil robot reaching through your laptop to personally steal your career. They're tools: very strange, very capable tools that can make experienced people dramatically faster in some situations while allowing inexperienced people to generate extremely convincing nonsense in others.
Learn them. Question them. Test them. Know what you're handing over, and know what you want to keep for yourself. For me, that means continuing to experiment while deliberately protecting some of the skills I don't want to lose: writing tests, reading code, working through logic, checking assumptions, and sometimes forcing myself to think about the problem before immediately summoning the robot.
Occasionally that also means opening those dusty university notes and remembering how we survived the before times.
puffs mysterious pipe again
We had Google. We had Stack Overflow. We had error messages. We had each other. And the elders? They had books. Terrifying.
Now we have something different: something enormously useful, sometimes unsettling, often impressive and occasionally dumb as a bag of USB cables.
So use the tools. Or don't. Draw your line. Move it when you learn something new. Ask questions. Stay curious. Do safe things. Do good things.
I will have no part in icky things.
And if we're going to have access to technology this powerful, maybe use some of that power to change the world for the better. Cure an illness or something, please. I'm rooting for you, sweet taters.
Affiliate Disclosure
Octivary may use affiliate links where available. Affiliate relationships will not determine how products are scored within the Choice Compass.