Weeds in a Garden
Opinions about AI are like weeds in a garden nobody's tending.
Everybody's got some, most didn't plant them on purpose, and half of them are just what blew in from the neighbor's yard. Say a claim enough times at enough dinner parties and it stops sounding like an opinion at all. It starts sounding like weather. Something everyone agrees is happening, nobody agrees on why, and everybody's got a coat for.
Some of these claims are true. Some are half-true wearing a nice suit. Some are just noise smart enough to get invited back.
I should admit my bias up front. Last year my team got disbanded, my boss got let go, and I got handed the bag. I spent some time after building (using AI) alone, without so much as an enterprise account to my name, before gaining a strong, competent team to build with.
None of what follows is theory I picked up from a think piece. It's what actually held up when I put my weight on it, and what quietly gave out.
Quick translation before we start. Most of what people call "AI" today is a large language model, or LLM: a system that's read an almost obscene amount of text and gotten frighteningly good at guessing the next word.
Not thinking. Guessing. Very, very well.
Keep that in your back pocket. It'll explain half the claims below.
"It's going to take all our jobs."
Verdict: False. But not nothing.
Picture water finding cracks in a sidewalk, not a tidal wave. AI isn't washing away whole professions overnight. It's seeping into the cracks: repetitive, low-judgment tasks like basic data entry and first-pass admin work.
Skilled trades stay largely safe. Try automating a plumber out of your crawlspace. White-collar work is murkier. It depends less on the industry and more on how much of the job is judgment versus repetition.
And there's a demographic wildcard nobody mentions enough. A huge wave of experienced workers is heading into retirement this decade. Somebody has to fill those seats. AI or not.
"AI is going to fix everything."
Verdict: False.
This is doom-and-gloom's cheerful cousin, and it's just as wrong. AI is a tool, not a miracle. It can speed up hard problems: drug discovery, climate modeling, tedious research.
But, it has no opinion on what we should do with that speed, and it can't fix things that are really about human incentives, politics, or who gets what. A sharper hammer still needs someone deciding what to build.
"AI is going to make things cheaper. Or more expensive."
Verdict: Genuinely TBD.
Depends where you're standing. Right now a lot of AI access is subsidized. Companies eating costs to get you hooked, the way a coffee shop hands out free samples.
Whether that turns into cheaper goods and services across the economy, or just a new expensive layer everyone has to pay for, is an open question with real arguments on both sides.
"AI is going to become as normal as smartphones."
Verdict: True. It's basically already happening.
Remember when "just Google it" turned into a normal sentence? Same arc, faster. AI is showing up in search bars, email, customer service chats, phone keyboards, often without anyone announcing it.
The technology doesn't need to be perfect to go everyday. It just needs to be useful enough that people stop noticing it's there.
"AI makes things up. Confidently."
Verdict: True.
This is AI's most quietly dangerous habit. It has a name: hallucination, when a model states something false with total conviction. It doesn't lie the way a person does. It has no idea it's wrong.
Think of a student who didn't study but interviews beautifully. The delivery is smooth. The content isn't reliable. That confidence is exactly what makes it risky. A hesitant wrong answer gets double-checked. A confident one usually doesn't.
"AI just reflects the biases it's trained on."
Verdict: True.
Every AI learns from training data: the mountain of writing, images, and information it studies before it ever answers a question. Whatever went into that pile shapes what comes out of it, blind spots included.
It's a mirror, not a moral compass. If the source material leans a certain way, the AI leans with it, unless someone actively corrects for it. This has shown up everywhere from hiring tools to image generators. It's one of the better-documented problems in the field.
"AI is going to become superintelligent any day now."
Verdict: Speculative.
Here we've drifted from fact into forecast. What people usually mean is AGI: artificial general intelligence, a hypothetical system that matches or beats humans at pretty much any intellectual task, not just narrow ones.
It doesn't exist yet, no matter how often it's discussed like it's parked around the corner. Serious researchers disagree, sharply, about whether current methods lead there or hit a wall that needs an entirely different idea. Anyone who claims to know the timeline is selling you something. Confidence, a book, a headline.
"You can always tell when something's AI-generated."
Verdict: False.
Detection tools exist. They're unreliable, especially for text. Picture a forger and an authenticator locked in a permanent arms race. Every time detection improves, generation adapts to slip past it. Treat any "AI detector" verdict as a hint, not a ruling.
How to Filter Through Piles of Cow Dung
The list above isn't just a fact-check, it's a template.
Sit with a handful of claims long enough, trace out where they're solid and where they wobble, and a pattern emerges in how good thinking about AI actually works.
A few rudiments worth keeping:
If you're standing with the crowd, that's a cue.
Old wisdom: when everyone agrees on something, check it twice. Not because crowds are always wrong, but because consensus tends to form before the thinking does. Someone says something quotable. It gets repeated. Eventually "repeated a lot" starts feeling the same as "true." AI conversation runs on this fuel. Ask someone why they believe their hot take, and more often than not you'll find a headline and a podcast played at double speed underneath it. Not a considered position.
The loudest voices on both ends are usually the least trustworthy.
There's a doomsday chorus certain AI ends the world, and a rose-colored chorus certain it fixes everything. Both are noise. Both are, frankly, a little lazy. A strong opinion is easier to hold than an accurate one. The people who actually know what they're talking about tend to say the least, because they've noticed how uncertain this moment really is. Confidence and correctness aren't the same thing. Right now they're often opposites.
The companies building this aren't your enemy or your savior. They're competitors, and that matters.
Tempting to cast AI companies as villains hoarding power or heroes ushering in utopia. Neither framing earns its keep. What's actually happening: a handful of companies fiercely competing over a tool they're still refining in public, racing each other for attention, adoption, relevance. That competitive pressure explains the hype, the pivots, the bold claims, far better than any grand conspiracy does.
The tool doesn't have values. The person holding it does.
AI doesn't decide how it's used. The person or company wielding it does. That's basically the alignment problem in miniature, the ongoing, unsolved effort to make sure what a system does matches what we actually want, not just what it's technically capable of. It puts the responsibility back on us. Get informed. Question your own assumptions as hard as you question the tool's. Look past the obvious use cases toward the strange, unconventional ones nobody's touched yet.
We are genuinely, uncomfortably early.
Worth saying plainly: the dust hasn't settled. It's easy to sound like a prophet right now and much harder to actually be right. That discomfort is exactly why the people who've thought hardest about this tend to talk the least about it. They're busy running u-experiments, informing themselves and otherwise doing ...
I felt this in my bones at an IT leadership summit in Washington, D.C., picking tools and wireframing a roadmap alongside people with a lot more seniority and a lot more certainty than me. I left half thrilled, half seasick. It felt like we were building a plane after we'd already jumped. Turns out that may well be the industry's current cruising altitude.
Where We Go From Here
None of that uncertainty means nothing interesting is coming. Quite the opposite.
Think dot-com boom: a lot of money, a lot of noise, a lot of companies that won't exist in five years. Some founders will have poured time, energy, and credibility into the wrong bets. That's how these cycles go.
But out of that same chaos came Google, Amazon, Facebook, companies that didn't just survive the shakeout, they defined the next twenty years. Good odds this cycle produces its own version: a handful of companies nobody's watching yet that end up mattering enormously. Maybe it's the first billion-dollar company run by essentially one person, leaning on AI agents instead of headcount. Maybe it's a dark horse like Sakana AI, precisely because today's leaders aren't guaranteed to be tomorrow's. Size isn't the advantage it used to be. I watched two employees quit within two months of each other and ended up with five hires and a real product in production by June. Small teams beat big committees. They always have.
So here's where the weeds come back in.
You can't pull every weed in a garden this size, and half the time you can't even tell a weed from a flower until it's had a season to show itself. Some of what's growing right now is going to turn out to be kudzu. Some of it's going to turn out to be dinner. Nobody, including the people funding this stuff at nosebleed valuations, actually knows which is which yet.
What you can do is get your hands dirty instead of standing at the fence. Go pull a few weeds yourself. Try the tool. Break it. Watch it hallucinate something with a straight face and learn what that feels like firsthand, so the next confident wrong answer doesn't sail past you. That's worth more than any hot take, mine included.
The honest takeaway isn't "AI is good" or "AI is bad." It's that we're standing in a garden nobody's finished planting, and the people who'll understand it best aren't the ones shouting from either end of the fence. They're the ones on their knees in the dirt, figuring out what's actually growing.