After twenty years of testing software, I’ve come to a simple, devastating conclusion:
all software is broken.
Not “occasionally buggy.” Not “sometimes unreliable.” Broken — permanently, irrevocably, gloriously.
Some systems simply disguise it better than others, usually with error messages so cryptic they could double as avant-garde poetry. “Unexpected error occurred.” Unexpected by whom, exactly? Certainly not me.
⸻
Myth #1: “You can’t trust AI. You can trust human code.”
This is a delightful myth, roughly equivalent to preferring to be bitten by a crocodile rather than a shark because crocodiles wear nicer shoes.
Human code is no more trustworthy than a chocolate teapot. It just takes longer to melt. AI delivers the same mess in ten seconds, and then rewrites it immediately when you point out the mess. Humans, by contrast, schedule a retro, blame QA, and write a twelve-slide deck about “communication gaps.”
⸻
Myth #2: “But it hallucinates.”
Yes. It hallucinates.
So does every developer when handed a vague requirement.
The difference is that when Bob from accounting hallucinates the wrong feature into production, we don’t call it hallucination. We call it “scope creep,” nod gravely in meetings, and then buy donuts to soften the blow.
⸻
Myth #3: “It isn’t perfect.”
Correct. Neither are we.
The crucial distinction: AI is imperfect faster. It breaks in seconds, learns in seconds, and patches itself before you’ve even finished explaining the bug report. Humans, meanwhile, will spend a week debating tabs versus spaces before admitting the unit tests don’t pass.
⸻
Myth #4: “AI takes jobs away.”
So did compilers, linters, CI pipelines, cloud services, and version control. Funny how nobody mourned when Git saved us from passing around zip files named final_final_v7.zip.
⸻
Myth #5: “AI doesn’t understand context.”
Neither do humans. I’ve seen entire sprints burned down because someone misread a Jira ticket. At least AI admits confusion instead of nodding politely in stand-up.
⸻
Myth #6: “AI can’t innovate.”
Neither can most corporate projects. Ninety percent of dev work is CRUD, with a side of boilerplate. If you want creativity, perhaps stop copy-pasting Stack Overflow answers with a fresh coat of comments.
⸻
Myth #7: “It’s just predicting words, not real intelligence.”
Yes — and your brain is just predicting neurons, which you call “intuition.” The only difference is that AI doesn’t need coffee first.
⸻
Myth #8: “It can’t be held accountable.”
Neither can Jenkins when it blows up at 2 AM. Accountability has always belonged to the people using the tools.
⸻
Myth #9: “It’s biased.”
So are humans — in fact, we invented bias. At least with AI, you can measure and correct it instead of pretending your decisions were neutral all along.
⸻
Myth #10: “You can’t debug AI.”
You can’t debug most legacy enterprise systems either, yet somehow we trust them to run payroll.
⸻
Myth #11: “It doesn’t care.”
Neither does the bug tracker, the CI pipeline, or your operating system kernel. Caring is not, nor has it ever been, a software requirement.
⸻
The Cosmic Joke
The grand secret is this: software has always been a sprawling, magnificent disaster.
AI doesn’t change that. It simply speeds the disaster up — and then resolves it in record time. It doesn’t eliminate imperfection; it makes imperfection iterate at lightspeed.
So yes, AI is imperfect. But so are we.
The only real difference is that AI won’t insist it was QA’s fault.
Top comments (2)
Thank you for this 🫶 Everything you said is hilarious only because it's true! I was beginning to think I was the only one aware of this situation. Glad to know I was wrong 🤣
Appreciate you, Ashley. You’re definitely not the only one who sees it. A lot of the pushback against AI feels like fear in disguise. Change always looks chaotic until you realize it’s just evolution moving faster than we expected. The ones who adapt will thrive. The rest will be busy making slides about why they didn’t.