Your AI is shit and useless for a major part of the population.

Okay, well, yeah, another frontier coming. Yeah, it costs, it costs how much? 50, 60, 100 bucks.
Oh yeah, very expensive.
And, well, who exactly are you expecting to use that?
Who can afford such massive spending on that?

The rich do not care

The people in those companies preach that it is for everyone, that it benefits humanity or blah – and then make it insanely expensive. Therefore allowing only the already rich people, in those rich countries, to find it useful. And thus creating and helping to support the endless cycle of the rich getting richer.

Yes, because the rich are able to afford that. The rich are getting richer because they can automate more – and not only automate more, but also get more and scale. And even if not scale, then at least a much better quality of life through reduced work hours.

Justified at work, unusable at home

Okay, well, yes. Assume you could use an expensive model for work. Yeah.
You can justify it, well, yes – if you pay 5-10 bucks for a task and you earn 20 an hour, you could justify that this is helping you somehow. Which I still don’t really get – how are you benefiting here, when you are unlikely to get twice the salary boost just because you use some tool. And most of you have to pay to use it anyway, from your own wallet.

But I’m trying to say that all these big fancy models are justifiable and useful only inside business and commerce. In there they can justify and use them – but not in your personal life.

You cannot afford all this big fancy shit for day-to-day interactions. You cannot afford to use it for your own personal projects, to automate your life away from the work.

Your hours do not change

Well, you see how that happens. In your workday, you still work 8 hours. You’re getting a productivity boost because of that. Has your pay increased? Haha, no – you still work the same hours, possibly even more, because now you have more pressure to use the tools more actively and all that. Whether you are paying for it directly from your own wallet, for all that API, or the employer is paying – you are kinda… maybe you get some relief from better tasks, but not much.
Yes, all these AI programs are gonna help you work – and like, work the fuck you work for.

And this is not even a new trick. The productivity-pay gap is documented for decades before AI – worker output grew faster than wages in basically every rich economy. Your employer capturing your AI-assisted output without paying you more is not paranoia, it is the default.

And then you come home, and you have to use them, because you cannot afford it. You cannot afford this fancy model that costs 50 bucks – you can only afford the one that costs less, and that one is twice as stupid.
And yet it is exactly in your personal life where the time from routine, from chores and all that, is supposed to be saved – and it is not actually saved.

You see how shit it is. At work, your time is not being saved, because you still work the same hours – but your productivity is getting boosted, and you don’t get an increase in salary for it.
And then you come home, and there you cannot benefit from all this frontier crap – the crap that could actually make your life twice as easy, like cutting your bullshit maintenance hours from 3 hours a week to 50 minutes or something.

And yet you are expected to feel the benefit of all that?

The rational split

Well, if anything, if you were paying from your own wallet, then the rational choice would be this. On your actual work, use the dump model as much as you want – like, what’s the matter? Why would you do more than the same stuff, when the pay is roughly the same, huh?
Then actually save up and use the expensive model in your day-to-day life, where your life actually matters.

And someone will say – on a subscription you just use the same plan for work and life. Sure. Even then you have to either allocate the major part of it to work, so you don’t get empty there, or ration yourself anyway. The result is the same. Your life is getting the deficit here.

And yes, of course, the models get cheaper every year, 10x or whatever. But who needs the old stuff? Why would you even want to use anything that is a year old? It is the same principle as everything here. You have to use cheap and stupid. Because the smart ones are reserved for work, to justify the spend there.

So, well, yeah, all of your crap is bullshit if it is not cheap. If the poor person in Tanzania cannot use it from day one without sacrificing half their food budget.

Where I landed

And always the increases and improvements are coming, day after day, one after another – and yet for 3 months now I have been using DeepSeek 4 flash.
Because this is the place where I can stop looking at the cost. At all. Just fucking at all. Not to mention it is just excellent, all in one context window, so I can just stop looking at that number at all. Really, this is the only cutie that gives me peace of mind of not even looking at those numbers. I don’t care how much it costs, I don’t much care how much context it took – because, well, why the fuck? It is negligible.
I throw 30-50 bucks a month on the API and just don’t think about it anymore.

Freed time is zero-sum

I want to point out the hypocrisy. Everyone promises it will help everyone, that it will free your time. But the freeing was supposed to happen at work – and it does not actually happen, because you still work the same hours. And at home it is not happening either, or happens at a much reduced rate. So you’re not getting the real gain for your quality of life.

Because freed time is zero-sum. Your employer and you are splitting the same hours, and the employer captures the freed time first – as extra output, not extra leisure. The slack gets absorbed instantly. What is left for you at a job is nothing.

Sure, any work use is a real benefit – freelancer, solo entrepreneur, whatever. But even there it is still a factor of cost versus personal. You are still paying out of your own pocket. And even if you are a freelancer, you still have a fucking budget, you just cannot spend as much as you want. You cannot spend equally on your work – which translates directly to revenue – and on something that is the other part of your life. One will always start to calculate, and it’s all scuffed in the end.

That is the general rule here, actually. Your money only counts if it is legible as revenue or wage. Spend it on work – it translates to revenue, justified. Spend it on your life – no receipt, no measurable return, it loses the budget competition every time. The pricing structure itself pushes all usage toward commerce and away from your personal life. Regardless of who is paying.

Like eating better at work

Some things that you buy benefit your life and your work at the same time – they are shared. You get a nice car to drive to work, and you can drive it anywhere. You get nice clothes, nice hardware and all that.
You buy one tool that improves your life both at work and in general, used at both sides.

Not here! Here you have to pay for stuff for work, and it is better than what you get for your life.

You know how that sounds? Like the fucking food! Imagine if you were eating better at work than at home. Like, really – at work it would be worth spending a little more for good quality food, if that could give you performance, when it is justified.
But not at home. At home you would be eating cheaper crap. Thankfully the food does not work like that. Else you would be living in a nightmare already – where a good employer means you eat well.
Maybe more like some mind improving drug. The analogy is still the same… hold on… it is!
The software is literally a mind capability improving drug, because if it is an extension of your mind, carrying out your will into existence – then whoever has it better, has it all.

The lesson

Now, what is the lesson out of all this?
First, to cut through the bullshit that these companies try to sell you, that AI is gonna improve your life and blah blah blah, while they are collecting the margins.

We have a very clear alternative, where the weights are open and you pay only for inference.
There are no margins there, so the price can be true to the cost actually required.
Both – proprietary and open – run on the same hardware. And yet the proprietary models, on the same GPU, suddenly cost 50 bucks, while the open ones cost only the raw allocation of GPUs and memory. Because that is all you pay for: the GPU allocation as it is. It does not matter which model you run – only how much space the model takes. If it needs two GPUs, you pay for two GPUs – the cost of maintaining them, the datacenter, and a small margin for the operator. Regardless of what model actually runs in there.
That is fair, and that cost stays more or less constant – software improves, gpus stay the same. Each month we have smarter stuff running on the same hardware.

Second, just support everything open, so stuff can be cheap and accessible.

Third – when you are at work and being provided access to subscriptions or plans, you have to demand that those subscriptions and usage are available in your free time too. So the cost of that subsidy is applied to your life as well. Demand what is applied, and try to convince your employer that time savings in your private life are as important for productivity as in the workplace.

And keep those fuckers accountable – for what they promise you versus how much actually you can integrate into your life.