Sylergy Logo

Everybody Knows AI Now

[ Sylergy Team | 2026-05-26 ]
Navigation | Blog > Everybody Knows AI Now

In a time when everyone has something to say about AI, you probably keep scrolling past certain posts without reading them. You can guess what to expect even before the text loads. Headlines with colons and promises like I tested 17 different AI tools so you don’t have to — often with a promise to do something 10x better. Carousel formats. Hooks that reel you in without giving away what they’re about. “Threads” that claim to be threads but basically rewrite the same sentence eight times with as much whitespace between them as possible.

The person who wrote it has never shipped a line of code that someone else had to maintain. They’ll explain transformers to you anyway.

This isn’t meant in a nasty way. I mean it’s a feature of our time — but those who could call it out directly, don’t. The people with the greatest influence on how billions of technically inexperienced people understand artificial intelligence are, disproportionately, people who never built any of it. They don’t have the experience of a model reliably delivering faulty output and you having to fix errors until 2 a.m. They don’t have to explain to the finance department why a human has to stay in the loop for these “agents.” They found a good wave, they’re riding it, and they’re incredibly good at surfing. Surfing itself is the art.

Person on a glowing data wave holding a phone, while a dimly lit office with engineers works in the background.

The bait that doesn’t even work

Here’s the part that should bother you more than the influencers do.

Those clickbait headlines — colons, curiosity gaps, you won’t believe this — are legitimate, studied techniques. Linguists call them information gaps. Good clickbait names a thing and then deliberately withholds it, creating a small itch of not-knowing that only a click will scratch. Kate Scott’s 2021 work in the Journal of Pragmatics describes it precisely — headlines that get you to “construct new conceptual files” out of the headline’s terms “while providing little or no content for those files.” (Scott, 2021) This style isn’t accidental. It’s deliberate.

And here’s something purveyors of “viral hooks” courses would never admit: it doesn’t reliably work. A controlled experiment from Penn State’s PIKE lab, published at CHI 2021, found that clickbait did not generate more curiosity than plain headlines. The simpler headlines often won. The paper’s own conclusion is that it “exposes the unreliability of clickbait.” (Le et al., CHI 2021)

There’s a further twist: the platforms themselves are supposed to be against this. Meta has demoted what it calls “engagement bait” since December 2017 — posts that “seek to take advantage of our News Feed algorithm by boosting engagement in order to get greater reach,” with “stricter demotions” for repeat offenders. Their own stated reason: people told them they hate it. (Meta, 2017; Meta Transparency Center)

Wall of identical social posts with colon headlines fading into a horizon of copies.

So sit with the contradiction. The bait is researched, but unreliable. The platforms claim to punish it. And yet your feed is full of it. This isn’t a story about a clever algorithm rewarding clever hooks. It’s a problem of sheer volume — a thousand people posting the same format until something resonates somewhere. Even when the algorithm demotes you, you can still be everywhere when supply is effectively infinite. They’re not winning because the hook is good. They’re winning because there are so many of them, and shipping a hot take costs nothing while shipping software costs everything.

Where the bill actually comes due

If this were just noise, I’d let it go. Noise is the tax you pay for an open internet, and I’ll pay it.

It isn’t just noise. The downstream costs are measurable, and they’re landing on the people who actually build.

Stack Overflow ran its 2025 developer survey across roughly 49,000 engineers in 166 countries. Adoption of AI tools is up — 84% now use or plan to use them. And yet trust is at an all-time low. Positive sentiment dropped from over 70% in 2023–24 to about 60% in 2025. More developers now actively distrust AI accuracy (46%, up from 31% the year before) than trust it (33%). Just over 3% “highly trust” the output. (Stack Overflow 2025)

Why the collapse? The survey is blunt about it: the number-one frustration, cited by 66% of developers, is “AI solutions that are almost right, but not quite.” Nearly half say debugging AI-generated code takes longer, not shorter. (Stack Overflow 2025) Stack Overflow’s CEO put the misinformation risk in plain words: “AI is a powerful tool, but it has significant risks of misinformation.” (press release)

Developer staring at code that is almost correct, with a fraying trust meter.

Now connect the wires. The gap between what AI can actually do and what people believe it can do is precisely the gap the hype machine widens for a living. Researchers have a name for this too. Savannah Thais, in a peer-reviewed 2024 paper, defines AI hype as “any non empirically or rigorously theoretically supported performance claims, capability narratives, or system descriptions” — and catalogs the harms of overstating capability: “subjecting the public to dangerous or harmful technology, artificially restricting research and development directions, and enabling misguided or detrimental policy.” (Thais, AIES 2024) She’s pointing at researchers and media, but the influencer economy is the same machine with the brakes removed.

And at the industrial end of the spectrum, the slop is no longer metaphor. NewsGuard has now catalogued more than 3,006 AI-generated content-farm sites, growing by hundreds a month, churning out dozens of articles a day with so little oversight that you can find the chatbot error messages left in the text. They are, in NewsGuard’s words, “often the originators of false claims” about brands, public health, and political figures. (NewsGuard) That’s the floor this content economy is built on. The influencer who can’t explain an embedding and the content farm that can’t spell-check itself are drinking from the same well.

Factory conveyor belt mass-producing identical AI articles with visible error glitches.

The honest part, because you’ll get it thrown at you anyway

Here’s where I have to be fair, because the cheap version of this argument is wrong and you deserve the better one.

Credentials are not the test. A degree isn’t a personality, and it definitely doesn’t guarantee you’re right. Some of the sharpest, most-followed voices in AI are credentialed to the teeth. Arvind Narayanan is a Princeton CS professor; his AI Snake Oil co-author Sayash Kapoor was a software engineer at Facebook before his PhD. (Princeton University Press) They’re exactly the kind of people who should have a big audience — and they do.

The strongest version of the counter-argument comes from inside the house. Ethan Mollick — Wharton professor, ~400,000 LinkedIn followers, about as “influencer” as a tenured academic gets — makes the point I have to concede: “I meet a lot of very smart AI critics who never seriously try to make AI work for them by spending a couple of hours with a frontier model working on hard tasks. As a result they are often misinformed.” His claim is that experience, not credentials and not raw IQ, produces the sharper take. (Mollick, 2025)

Split image: polished influencer setup versus worn hands on a keyboard.

He’s right. That dismantles the lazy version of my complaint — but sharpens the real one.

Because the dividing line was never the degree. It’s the calluses. What makes someone worth listening to about AI isn’t a CS degree, a verified badge, or a follower count — it’s whether they’ve actually put their hands on the thing and felt it break. Mollick’s “smart critics who never tried it” and the influencer who’s never shipped a thing are the same failure mode in different jackets: confident commentary with no contact with the surface. One just happens to have a curiosity-gap headline attached.

So what do you actually do

You can’t fix the supply. There will always be more people willing to post a hot take than to debug a flaky pipeline, because one is free and one is not.

What you can fix is your filter. Stop using reach as a proxy for knowing. Follower count tells you someone surfs well; it tells you nothing about the water. Before you trust a take on what AI can do, ask the only question that matters: have you actually built the thing, and did it break on you? If the answer is a carousel, keep scrolling.

Lone builder working under a desk lamp while distant screens flash with notifications.

The airwaves belong to the loud. They always have. But the people quietly shipping the real systems — who know exactly where the model lies because it lied to them at 2 a.m. on a Tuesday — are still here, still building, mostly not posting about it. Go find them. They’re terrible at hooks. That’s how you know.

Sources

Contact

Let's streamline your operations.

Tell us a bit about your workflow and choose how you’d like to continue.

1

Save 10–20+ hours weekly

Auto-pilot redundant data copy/pasting processes automatically.

2

Reduce human error

Verify data constraints and system integrations automatically.

3

Built for how you operate

Tailored platforms that fit your staff's actual workflow rules.

4

Secure, scalable, and reliable

Modern client portals backed by strict authorization structures.

How would you like to continue?

Your qualification call will be led by our AI assistant and usually takes 10–15 minutes.

  • We’ll discuss your workflow, bottlenecks, and goals
  • No pressure or sales script
  • If there’s a fit, we’ll line up the right specialist

Prefer to choose a time later? Schedule the same qualification call for a slot that suits you.