Have a look for example what has happened with Frontiermath: https://epoch.ai/frontiermath

GPT-4.1 scored 0%, while GPT-5.2 (Pro) scored 29,2%. That looks like an improvement to me. Especially because those are research-level math problems and you can't find solutions to them from the Internet. And research-level problem means that if you are a good mathematician, you still can't solve those problems unless it is in your specific field.

It is not grunt work that teaches you how to work. I personally started programming simply by doing my own projects. It was no grunt work, I just did what I could and if I didn't know how to do something, I tried to learn how to do it. When I joined the work force, I pretty much continued to do my own projects again. I made a lot of mistakes and I learned from those.

I think we can easily let go of the grunt work, if we just offer the kids a place where they can do those big projects in an environment where they can make mistakes and where they can ask for advice. If they can't play with the real thing, let them play in simulation.

I just checked from two different operators how you autenticate eSIM. They don't send you an SMS, instead you login to their webpage using e.g. your bank account to identify yourself. Then you scan QR code and that is about it. (You will need a WIFI connection.)

So AFAIK there is nothing in the eSIM that mandates the usage of SMS. That is probably just operator specific way to operate.

No. Lump of labor is saying that the amount of work is fixed and I am not claiming that. What I am saying is that there is an upper limit of how much goods a single person will need in specific area. For example most people are quite happy with owning just one washing machine per family, which usually lasts about 8 years. So if we assume population of 8 billion, family size of 4, we need about 250 million consumer level washing machines produced per year. There is no point making much more of them, because no-one would want to buy them. And every time you increase worker efficiency with automation, you need less workers when making washing machines.

Counter argument is that there is unlimited amount of different machines you can make and sell, like TVs or 3D glasses, so workers from washing machine company can just build TVs. But you will still have limited amount of time, how much you can spend with those devices. Sure, you can buy a TV and just let it be unused, but I think people will have limit of how much unnecessary stuff they are willing to buy.

> First off, we don't know if AI and robotics are going to cause even field specific unemployment.

We can be pretty sure that it will happen, because:
1. For most goods, there is a limit of how much of it is needed.
2. Automation will increase efficiency, so less workers will produce more.
3. Required amount of workers for certain good = [required amount of goods] / [amount of goods that single worker can produce with automation]
4. Even if we can invent new products, like hologram games, those products almost always compete with existing products. E.g. hologram games would just take customers from video games and cause unemployment there.
5. In some fields there will be temporarily need for more workers. E.g. health care should be one of these fields as demand increases because the amount of old people increases. But that is only temporary. In the long run, automation will cause unemployment.

I agree with you that retraining won't help much.

What you describe is a software architecture problem. No decent software architect would allow code like that. Most likely the project doesn't even have one, or they have one named as a software architect, but in reality that person is just a normal developer. I have seen that happening multiple times. It is extremely important to pick a good architect for the project. That alone makes determines whether the project costs 100 million or 10 million.

I have been in agile projects that don't have such problems, so I don't think that agile is a problem, nor did you identify anything in your post that would suggest so. But agile can not save a project that is run by incompetent people. You have to have a competent person in charge or you will fail, no matter what methods you use. I have been in water fall projects that manage developers via tickets like you described. It is not agile specific thing, it is a project leadership thing.

I estimated that if I would shave half of the memory usage out of the application, I would have to do that in 8 seconds, including testing to make it worth the cost.

That is not true. Electricity in Sweden and Finland costs about 5 eurocents per kWH. That is pretty much as low as price can get without renewables. Well, in theory you could get the same price with 100% nuclear, but not lower anyway. Currently electricity costs about 0.4 eurocents per kWH in Finland, because wind production is high. And for the next 24 hours the price will be lower than that at all times.

There would be a chaos yes, but airlines would not implode, they would just try to manage in any way they can. Most obvious way to manage the chaos is to raise the ticket price and make less flights, possibly fire some staff.

> people that simply want to use MS Office have no interest in learning something new

I know people who have used MS Office with an external calculator to count the sum of values in their spreadsheet. So I think you are more right than you think.

With printers, I just bought a network printer. There was not much of a price difference and now that I have it, it works pretty much from any OS.

I don't know if I had to do this, but I like to make my work easier by planning ahead.

I am a professional with decades of experience. I think that AI is best at prototyping. If I have an idea, I can ask AI to write a program that does that. That program does not work properly, it is full of bugs, but it will instantly tell me a lot of things. Most valuable thing it can tell me is that "this idea does not work". That alone can save hours of my work. It will also reveal things that I didn't even think about, but which are essential for the idea to work.

Because the code is just used as a prototype, bugs and quality doesn't matter. But because I have that prototype, I can write the actual program with much more accurate plans, making the actual code much better in quality.

Making an AI like that is not hard. Making an AI that can fold proteins is hard. Guess which country managed to do that? (Answer is United Kingdom). Let me know when China makes an AI that can win a Nobel.

Microsoft has not been on the frontier in the AI race, so how would he know what is costs?

I think that pretty much all memory modules have spiked. I just checked Kingston (consumer) DDR4 16 GB memory module, which price has gone from 42 euros to 131 euros within 6 months. Kingston buys DRAM chips from Micron, SK Hynix, Powerchip, Nanya and others.