Open almost any software tool or app today, including several crypto apps, and you will be greeted by the exact same obnoxious design cliché: a little gradient sparkle icon or a button labeled 'Ask AI'.

Tech companies are desperate to shove generative AI into everything.

It's like duct tape for unfinished products.

But these buttons are mostly pointless.

It's a corporate box-checking exercise

Mostly, these buttons exist for one reason only: marketing.

Executives are keen to tell shareholders: "look, we have integrated AI".

A good example is the AI button integrated into apps that require customer support.

In general, these AI buttons are great at summarising the key points of the fine print, but, with exceptions, they can't get anything done.

I always end up telling it "talk to agent" incessantly, until it gives in.

And there's another major problem.

The hallucination problem

AI hallucinates all the time.

I use AI every day to proofread my articles (and by the way, it happens very rarely, but it also misses typos sometimes) and it gets stuff wrong all the time.

I remember when I had to write about the new foldable iPhone and had to tell AI that yes, the new iPhone had indeed just come out.

Large Language Models do not understand logic, truth, or context.

They predict the next most likely word based on probability.

By their own admissions (by which I mean, by the admission of the engineers that created them) they're designed to sound confident.

Try any AI, ask it: "are you designed to sound confident?"

It'll say yes.

They invent fake statistics, misattribute quotes, and hallucinate entirely non-existent facts if they have to.

Companies should stop trying to automate thought.

Software used to be about making tools more responsive, reliable, and efficient.

Now, it is about bolting nonsensical parrots onto interfaces just to keep the hype train moving.