And honestly I don't give a fuck if it's popular or not. It's going to make a lot of kids' lives better. That's the point of government. Not an opinion poll.

The leak that turned out to be right about it happening had some weird pre-payment scheme based on estimates of next year's mileage. That part might have been nonsense but for a government so keen on cutting red-tape they have proven to love their complicated bureaucratic rules and systems.

Making ICE cars less of a bureaucratic hassle than EV is probably not helpful where people don't expect to do a huge number of miles.

Possibly more complex than that. Reflecting on my own relatives I think it's more strongly a generational thing with slower discarding of the habit associated with class or wealth. Decline associated with when radios/TV first made it into homes perhaps?

Suspect it's not really measurable without access to something that would automatically be a better choice than what it was being used to measure.

Look at the media wildly speculating that milkshakes and lattes contain milk and sugar! So reckless!

I try to respect the hard work and skill that goes into polling but when so many questions are "Hello, what do you think about this thing you know fuck all about and have never probably thought about before me asking you?" and then presenting the results as some great insight makes it difficult.

Long lead times + FOMO I suspect.

If I need to order GPUs long before the building is ready, then any delay in that construction gives me a warehouse full of expensive GPUs. Even then when I expect delays with my other projects, do I want to risk my place in the queue by cancelling?

We should also tax advertising with 'enhanced' rates for risky-but-legal things like gambling or alcohol.

Have we all memory holed the RailTrack/Network Rail saga?

Can you extend this to other domains? For writing I think it's clear that poets and novelists are Carvers, journalists are Carpenters, but I feel that beyond that it might get a bit contentious. 9/end

Construction projects however are all about the ways you can join things together and the quality and consistency of training information will have been much, much stronger - and probably over a smaller range of programming languages. 8/

Carpentry type problems they will have been trained on but there will be many subtly conflicting bits of training data as every carpentry problem and solution is a bit different. 7/

Essentially by definition Carving type projects are ones the AI haven't been trained on as they don't exist yet. There might be some similar, but if they were very similar it would be Carpentry not Carving. 6/

General trend over the last few decades has been for Construction to become a bigger and bigger share of what sort of coding gets done.

There are a number of implications of this model, but the one I'm thinking about now is AI. 5/

* 'Construction' - Building large complex applications largely by attaching pre-existing components together. Lots of frameworks, scripting languages. Lots of reuse of components. 4/

* 'Carpentry' - Where you need to build something custom but can rely more on experience of similar projects and existing components. Code generators, data processing, device drivers, OS libraries - big on reuse of experience 3/

* 'Carving' - Where you have sense of what you need but have to explore the problem space. Research, innovation and low-level code for bootstrapping new HW (or virtual HW) - big on reuse of concepts. 2/

Prices won't drop until the banks stop lending because they have hit a regulatory limit or have done a Northern Rock. If incomes remain high enough for deposits and mortgage payments then banks can keep lending longer than builders can keep building.