Rebranding is not a feature.
TechCrunch reported something that should be taped to every product manager’s monitor: only 2% of consumers are buying what’s now being called “AI, call it Super Intelligence.” Two percent. That’s the number after years of demos, keynotes, funding rounds, and now a federal renaming ceremony.
If you missed the naming part, here’s the short version. President Trump announced that the U.S. government will officially refer to artificial intelligence as “Super Intelligence,” describing it as much bigger than the internet. Several major American tech company leaders signed a White House accord alongside the announcement. The order covers terminology used by the U.S. executive branch — it doesn’t rename AI worldwide, and existing laws and previously issued documents aren’t automatically changed.
So we got a new label and a signing ceremony. Consumers got the same products they already weren’t buying.
What 2% actually tells a builder
I build bots for a living. When I read a number like 2%, I don’t read it as a verdict on the technology. I read it as a verdict on the pitch.
Most people have now used something marketed as AI. They’ve asked a support bot a question and gotten a link to a FAQ page they already read. They’ve tried a writing assistant that produced text they had to rewrite anyway. They’ve watched a smart speaker mishear the same command three times. Their skepticism isn’t ideological. It’s experiential.
Calling that experience “Super Intelligence” doesn’t close the gap between what was promised and what showed up. If anything, a bigger word makes the gap more visible. A tool that fails at scheduling a dentist appointment is now failing under a grander name.
The naming instinct is a tell
I’ve seen this pattern inside engineering teams, not just governments. When a project isn’t landing, there’s a temptation to rename the thing. New codename, new branding, new slide deck, same unresolved latency problem. The rename feels like progress because it’s fast and it’s cheap and it produces something you can show people on a Tuesday.
Shipping a bot that reliably completes a task is slow and expensive and produces something you can only show people once it works. Guess which one gets prioritized when attention is the scarce resource.
The 2% figure is what happens when the naming instinct outruns the shipping instinct for long enough that users stop listening to names at all.
What I’d build for the other 98%
If 98% of consumers aren’t buying, that’s not a market to lecture. It’s a market to serve differently. A few things I’ve changed in my own work, for whatever it’s worth to yours:
- Pick one task and finish it. A bot that does one thing end to end beats a bot that does nine things halfway. Users forgive narrow scope. They don’t forgive unfinished.
- Stop saying “AI” in the interface. Say what it does. “Drafts your reply.” “Finds the invoice.” Nobody buys a category. They buy an outcome.
- Make failure legible. When the bot can’t do something, say so immediately and hand off cleanly. Half the distrust I see comes from systems that confidently do the wrong thing instead of admitting the limit.
- Measure completion, not engagement. Time spent talking to your bot is usually a bug. Tasks completed per session is the metric that tracks whether anyone would pay.
- Earn the retry. Most users give a new tool one or two attempts. Your onboarding path needs to produce a real win inside those attempts or you’ve spent your entire budget of goodwill.
Why this is good news for small builders
A saturated hype cycle with a 2% conversion rate is a strange kind of opportunity. The branding layer is maxed out. There’s no bigger adjective left — we’ve already arrived at “Super Intelligence” and “bigger than the internet.” Whatever comes next has to compete on whether it works.
That favors people who write code over people who write press releases. If the marketing ceiling has been hit, the remaining differentiator is reliability, and reliability is something a two-person team can actually win on. You don’t need a White House accord to make a bot that correctly parses a PDF and files it in the right folder every single time.
The part worth keeping
I’m not cynical about the underlying technology. The models are genuinely useful and I use them daily in things I ship. My complaint is narrow: the gap between the vocabulary and the delivered experience is now wide enough that consumers have priced it in.
Closing that gap is unglamorous work. It looks like error handling, fallback logic, better prompts, tighter scope, and saying no to features. It does not look like a new name.
Two percent are buying the story. Build for the ninety-eight who want the thing to work, and you’ll have a product either way.
🕒 Published: