The problem with OpenAI saying the solution to AI taking all the jobs is to raise taxes on corporations and the rich to pay for benefits for the unemployed is that they literally spent millions of dollars getting Trump elected to do the exact opposite.
Google is right to dogfood to improve Gemini. The day they give up and use Claude internally, Gemini is dead.
To be clear it's not a "homework" project, it's a 1 hour interview where the candidate shows how they work to the interviewer. It's not "free work".
Now the fact that you're talking about a slop machine, I suspect you haven't tried them.
2. It's stupid to do code interviews on a whiteboard without access to Google/StackOverflow for a job you'll have access to those tools at will. Similarly, in companies where AI coding tools are widespread it makes no sense to forbid them at code interviews.
1. I've done a lot of interviews in my career, and I know some candidates can bullshit their way if you just do talking, so I'm convinced that you need some kind of technical interview. We can discuss what a good tech interview is, but just talking is not enough
That's because solar and wind are not dispatchable, and that's why trying to get 100% of those 2 is problematic.
How is it horrible to create an interview setup that is as close as possible to how they will work on the job?
In my company we're experimenting with AI interviews now. Where the candidate must create a piece of software assisted with AI, showing his process to the interviewer.
That's cool, but you'd have to be careful not to turn that into being controlled by the one country that produces 80% of the solar panels in the world....
The point still stands even if it's not the only source. Taxes on gas are an important part, and if everyone drives an EV this part will have to be funded from another source.
In most countries road maintenance is paid by taxes on gas. As we transition to EVs, that don't pay that tax, we'll need to figure out how to pay for roads.
It does, heavier vehicles wear down the road faster
They might not have friends, but they have sycophants to spend time with and they enjoy it.
That's the thing: you don't get to the "super rich" level if you stop trying to get more as soon as you have enough to live in your man cave without working.
It doesn't matter for EVs.
They don't have the complex mechanics of an ICE so they're reliable no matter the brand.
Also, when I need to find my kids I just check the places that have outlets.
My kids developed a sixth sense for power outlets, leave them in a new place for a few minutes and they'll be able to tell you exactly how many there are and where.
I didn't notice you were a woman, but confirmation bias happens to everyone. If you think apps are broken because of AI, you're more likely to notice bugs that might have not bothered you before.
Are you sure software is less stable than before, maybe it's just confirmation bias?
Well I don't know, but it would certainly be a good idea to stop assuming tech CEOs are tech geniuses. There's no reason why they should be, and they often make silly claims about the tech their companies are building.
futurism.com/artificial-i...
D'après l'article son budget sortie c'est 400 à 500 euros par mois, une bouteille de champagne suivant l'endroit ça peut être dans les 100 euros... Pas vraiment besoin d'être milliardaire pour acheter une bouteille a 100 euros de temps en temps.
pour quelqu'un de seul, 3500€/mois logement payé même a Paris c'est confortable
As someone who lived through the cassettes golden age, I really don't understand the appeal today. They sound bad, they quickly get bad. You can get the experience of making your mix tape with minidiscs, and I'm sure you could replicate that physical media experience with a digital system.
You need to pick the right model for you question (or let your tool pick it for you).
Small and cheap models exist, you don't need a reasoning model to do a Google search.
Finally, smaller models like Qwen work locally on the kind of laptops SWE are already getting. Today they're far from being as good as frontier models, but as they get better we might see a switch to local models with no monthly spending.
(2) customers: they spend without counting because they don't want to miss on the new ways to work, but if they have to look as cost cutting they can like they do with cloud code. For example by favoring cheaper models like Sonnet over Opus for daily usage.
It still begs the question: if those companies lose money, is AI usage unsustainable? My take is that there are 2 things: (1) AI companies are using the bubble situation to grow a lot and invest. If they need to go "lean", they can and they will.
I also dispute the $2000 per month per developer, to reach that you need to do tokenmaxing or experiment a lot. So a company might have 1 or 2 engineers among their thousands spending that, but the average is closer to $200 to $300 per developer. Which is small compared to the total cost of a SWE.
