Shares get taxed on price, not on value.

Not yet. This would require some level of AI "agent" collaboration that is not really possible at this time without breaking basically everything. But model poisoning is already very feasible and probably done in practice, especially as the amount of poisoned training data needed does not seem to be large.

The dumb people are still there. Or have you somehow overlooked that a lot of people voted for this? Hence I have no clue what you are talking about when you claim I am ignoring things.

You need to take into account that the hardware does not keep forever. For this type of hardware, the errors probably start to become a problem after 5 years or so, because it is being run fast and hot. Hence hardware is not a sunk cost, but an ongoing cost.

Exactly my take as well. This is about making some people money, not about any real risk.

Definitively. Always insist to get the security of a known-good classical algorithm in addition. Relying on something "quantum safe" alone is pure insanity at this time.

My bet would be that breaking something like RSA 2048 with a QC is not possible in this universe. It is too small and has not enough remaining lifetime.

It is not old enough and may still fail with catastrophic weaknesses. The way to go for new products is to use hybrid encryption, where a successful attack requires breaking both a quantum safe algorithm and a classical (good) one. Or, if you can, stay classical, since Quantum Computers are very, very, very, very far removed from being able to break any real encryption. In fact, after more than 50 years of research, these "machines" can factor 29 currently (well, one could and that was with moderate cheating as it was not with the general algorithm, but with one specifically adapted to factor 29). For reference, that is 5 bit. Current RSA recommendations are 2048 bit or longer and 4096 for long-term, security. If the keep doubling these bits every 50 years (which is a really big "if"), RSA 2048 will be within reach in about 400 years.

Disable, yes. Uninstall, no. If it's pre-installed it's part of the system image, which is mounted read-only and protected with fs-crypt. Actually modifying that would require root access to remount it rw and to disable fs-crypt.

That would also, of course, completely destroy the Android security architecture, leaving you wide open to all sorts of attacks. If you want to do that, get an Android device that has an unlockable bootloader (e.g. Google Pixel), unlock it, then do whatever you like. And be sure not to hire any evil maids.

I'd recommend ML-KEM and ML-DSA instead.

I also use the Claude Code extension with VS Code. The inline diffs it provides are quite clumsy compared to Cursor's.

The integration is a little better in Cursor; the main difference being the in-line edit diffs. But I bounce back and forth between Claude Code and cursor, so I end up just using the git diff view to look at changes about 80% of the time, so it's not much better.

Honestly, my reason for using it is that I have separate Claude and Cursor token budgets -- though I set Cursor to use Claude so I'm using the same model both ways.

If your government breaking into your house and applying hardware-level attacks to scrape your secrets out of the RAM of your running computer is seriously part of your threat model, it's almost certainly very, very far from your biggest concern.

Also, you should probably consider turning your computer off.

Yes. And I hink the Brave filter is working quite well.

Indeed. MAGAs are hallucinating worse than the least capable LLMs. Quite an achievement, if not a positive one.