Why is everyone suddenly an "AI thought leader"?
The volume of AI evangelism is a signal of low trust, not high adoption. When software acts autonomously and its output is hard to verify, the only thing left to sell is belief. So people sell belief — loudly. A recent Ask HN thread captured the mood: feeds full of "you're a dinosaur if you're not doing this," with almost no working examples attached.
The pattern is familiar. During the 2021–2022 crypto peak, confidence marketing outran shipped utility, and Gartner's own hype-cycle framing predicts exactly this: a "peak of inflated expectations" precedes a "trough of disillusionment." AI is following the shape.
Here's the tell worth naming. Evangelism — persuasion aimed at getting you to believe a technology works — thrives where verification — the ability to check that it actually did the right thing — is missing. The louder the belief-selling, the weaker the checking underneath.
What the Ask HN thread actually shows
The complaint wasn't "AI is fake." It was that the content is insufferable: daily posts, recycled influencer takes ("Boris McAI said software engineering is dead — here's 5 reasons he's right"), and shame-based framing. Notice what's missing from that description: a demo, a before/after, a number. The posts assert transformation. They rarely show it.
Why does hype grow when the tool is autonomous?
Autonomy and evangelism feed each other. When a tool commits actions on its own — sends the email, updates the record, moves the money — you cannot easily inspect the reasoning before the fact. That gap gets filled with narrative. The result is a market that runs on confidence instead of evidence.
An autonomous agent is software given authority to take actions without a human approving each one. That design maximizes the surface area for hype, because:
- The work happens off-screen, so trust must be asserted rather than shown.
- Errors surface after they've already committed, making them costly to attribute.
- "It's magic, trust me" becomes the honest description of the user experience.
MIT's 2025 NANDA report on generative AI in the enterprise found that roughly 95% of organizations saw no measurable return from their generative-AI pilots. That is precisely the environment where evangelism flourishes: adoption is being talked about far faster than value is being produced.
What actually replaces the evangelist?
Verifiable software. The antidote to belief-selling is a system that shows its proposed action, waits for a human to approve it, and leaves a record. When operators can check, they stop needing to believe — and the thought-leader theater loses its audience.
This is Vantrow's guiding principle: propose, never commit. The software stages an action — a draft, a change, a recommendation — and a person decides. Concretely, governed software does three things the autonomous demo does not:
- Proposes, then waits. Nothing lands until a human approves it.
- Shows its work. The inputs and reasoning are inspectable before the action, not after.
- Leaves a trail. Every decision is attributable, so results can be measured instead of narrated.
Governed AI beats autonomous agents for the same reason audited books beat a confident CFO: you can verify the outcome. Once you can verify, the sales pitch stops being "believe me" and starts being "look."
A quick test for the next post you see
Ask one question: does this show a result I could check, or ask me to believe a claim I can't? Working software invites inspection. Evangelism discourages it. The distinction sorts most of your feed in about three seconds.
So is AI overhyped or underused?
Both, in different rooms. It's over-narrated on social feeds and under-deployed in production, and those two facts are connected. The hype is loud because the durable, governed use cases are quieter and harder — they require staging, approval, and audit rather than a screenshot and a bold claim.
The operating companies getting value aren't posting about it hourly. They're wiring AI into workflows where a human still holds the commit, and where every proposal is logged. That's less shareable. It's also the part that works.
FAQ
Why is AI content on LinkedIn so annoying right now?
Because much of it sells belief rather than showing results. When AI tools act autonomously and their output is hard to verify, marketing fills the gap with confidence, shame ("you'll be left behind"), and recycled influencer takes instead of demonstrable outcomes.
Is this just another hype cycle like crypto?
It rhymes with it. Gartner's hype-cycle model describes a peak of inflated expectations followed by a trough of disillusionment, and MIT's 2025 NANDA report found ~95% of enterprise generative-AI pilots showed no measurable return — the classic gap between talk and value.
What's the difference between an autonomous agent and governed AI?
An autonomous agent takes actions without per-step human approval. Governed AI proposes an action and waits for a human to approve it, keeping the work inspectable and the decision attributable. Governance replaces belief with verification.
How do I tell useful AI from evangelism?
Ask whether the post shows a result you could check or asks you to believe a claim you can't. Working software invites inspection; evangelism discourages it. Proposals you can audit beat promises you can only trust.
Does "propose, never commit" slow teams down?
It slows the commit, not the work. Staging an action for approval adds a review step but removes the far costlier cleanup of autonomous errors — and it produces a record you can measure, which is what turns pilots into results.