I want to build things, not "launch startups". If someone asks me about a minimal thing of what I build, I want to be able to speak for hours about it, in detail. I want to completely understand it. I'm an engineer, not an enterpreneur.
This whole concept of Software 3.0 takes away all the craft from the thing. It's souless.
- "How did you do it?"
- "Dunno, just prompted it LOL."
Sounds very dumb to me. Even if you make loads of money off of it.
That's why I also believe it will not be a "gateway drug to software development". That's just a profound misconception of what makes good software developers tick.
They are now trying to sell the idea of control. You can "put it in a leash". It's pathetic. I don't want control, I want to understand it. Internalize it.
One thing that never changed in software development is that _you get the payoff_ for the hard work. You solve the mystery. You figure out how it works, by yourself.
Anyone that looked a little close to AI knows that it doesn't work that way. It's a black box thing. You will never fully get the knowledge payoff.
I'm sure he's a great guy, but let's face it: He's there speaking those things because he made some popular videos in the past and they needed to renew their PR strategy.
Speaking of PR strategy...
> [...] augments [...]
Iron Man does not buy his metal suit at Costco. He builds it, in a cave, with a bunch of scraps. That's the magic of building things.
> I really don't think people understand how great things could be.
That's not the issue. I think it could be great, but they're being greedy and underestimating their best audience.
Instead, they're focusing on complete beginners, in the hopes that those beginners will generate enough monkey-bashing to train a model that can churn good quality "software 1.0" (which is the real deal, so far irreplaceable). I believe that's a mistake.
Have you tried building things with this stuff? I have, it's exciting, there's all kinds of new things to build. I haven't been as excited to build things for a while before this.
That's what their PR material seems to imply.
> Have you tried building things with this stuff?
Yes.
> it's exciting, there's all kinds of new things to build
Yes, and no. I don't want to pay to build, or be dependent on an API or model delivery. It means I want to build the juicy stuff, the models themselves. Generate synthetic data, experiment with different approaches for training, etc. That is way outside the reach of common people (it can't be done now, you need lots of money).
I also don't want to buy a fancy GPU so early. Even the most expensive ones are currently very weak to do anything of real value. I could run some inference or adjust some weights, but I would still be at the mercy of some company delivering me base models.
The prices are all inflated because of the hype, and it seems AI companies are synchronizing this inflation with the hardware companies. I'm betting on this pricing bubble to pop.
Regarding the stuff I can do with already pre-trained models, there's not much to learn. In its core, it's basically the same good old stitching of APIs (I've been doing this for decades). I tried using the AI IDEs, but didn't found much value in them. I'm sure they're great for some scenarios, but it's more of a gimmick to use developers to generate training data than anything else.
Regarding "agents" and stuff, I have zero interest in competing in this "user of model" market. It's so easy to do, that any new idea is flooded with attempts and saturates in mere weeks.
When the opportunity to build something _really_ cool and novel appears, I'll give it a real chance. The race between big tech competitors is killing those opportunities for now, only people hired by big AI companies get to do it, and I'm not one of them.