Rob Wiblin

The index puts GPT-5 and other recent models kind of right in the middle of that recent faster trend.

So my interpretation is that behind all of the noisy swings in sentiment, people suddenly becoming optimistic and then horribly pessimistic, AI has neither gone to the moon nor has it hit a wall.

Instead, it's just gradually but relentlessly been getting a bit more useful every month for years now.

My second argument just stems from my experience using these models.

Sometimes I hear people say that AI, it's just not useful.

People don't want it.

And I just want to ask them what the hell they are smoking.

I effectively use AI for hours a day now as a co-pilot and a thought partner.

It allows me to get up to speed on things quicker, to learn things faster, to put my finger on exactly the thing that I don't understand and get the answer to it.

It also helps me get drudge work done much more quickly than before so I can get onto more interesting stuff.

It's also, as so many people have found, it's a great place to get advice on health issues, financial planning questions, social situations, whatever thing you're mulling over in your head.

It recently helped me to diagnose, just as one example of tons that I'm sure people could give, helped me treat this kind of longstanding issue I had with my nose being congested all the time.

It figured out what the issue probably is and figured out what I should do in order to mostly get rid of it.

Now, to be sure, it's usually not as good or reliable as the best human expert on these things.

But most of us, most of the time, can't turn around and get the best doctor or financial advisor or therapist or whoever to talk to us for as long as we like for free, more or less.

I think this is a huge change from the end of 2024.

Back then, I was kind of struggling to figure out a practical way to make use of these models as impressive as they seemed on paper.

So sometimes I think maybe people's impressions are just way out of date if they decided they don't like AI and they stopped trying it out after that point and they're not willing to pay for the actual cutting edge models now that might keep them abreast of what AI is actually capable of.

I mean, I think that's the most generous explanation that I could give for this kind of attitude anyway.

Third, and this one is really important, earlier I emphasized that the growing compute and financial cost of the very frontier reasoning and agenting capabilities that exist anywhere.