First think about the practical, legal and PR challenges of putting self-driving vehicles on public roads.

Putting autonomous robots in amongst humans is that problem, on steroids.

Edit: there’s a reason why industrial robots all have a hazard zone marked around them, and some even have transparent safety screens or cages

This concept of making the best use of any available data to infer useful probabilities seems like a crude form of some ideas behind current LLMs, BUT they (Peters, Graham, Robinson and Woodhead) were doing this work 20 years earlier.

Bayes theorem dates back 250 years.

This is what engineering is all about. Applying what it long known to solve new problems, or old problems better.

LLMs are a triumph today because someone made a keystone observation

>why the stasi didn't build a large pit somewhere in a land fill, put all the records in it and burn the lot.

In hindsight, that would obviously have been the most efficient, but it needs trust and cooperation between regional Stasi offices to give up control of their records.

But The Stasi did not have foresight, was not efficient, did not have cooperation internally. (If they had all those attributes things may have gone differently. So it’s good they were still incompetent at the end)

What about modern devices (such as laptops) that have a fixed battery? They shouldn’t rely totally on software to force a hard power off because that’s not safe. If the software hangs you cannot power down

The EC (embedded controller) controls the hard power off on a modern laptop, not the OS at all. That's why you can still power down in the event of a device lockup (thankfully).

It's all fun and games until the EC locks up.

Have had it happen on the Essential Phone and my MIL's Chromebook (don't remember the model). At least with the Chromebook, you can open it with a screwdriver and pull the battery. Better than waiting however many days for the battery to fully die so the EC resets.

Some laptops have a battery reset button for cases like this. For example lenovo has a pinhole where you can press a tiny button using a paperclip.

https://devblogs.microsoft.com/oldnewthing/20250107-00/?p=11...

This is legally dangerous for the company because those words have generally well understood meanings, so a lawyer could argue that they promise a certain level of service. It’s safer to use vague words to describe the product variants.

Windows used to have a method to map out bad memory areas with bcdedit. Did that ever work? Does it still work?

With the increase of soldered RAM it must be more important than it used to be ten years ago.

Just search for “bcdedit and ‘bad memory’” for more documentation

>the complete obliteration of the basic assumption that ‘someone wrote this so it is at least worth a cursory glance’.

The OP is not disputing that a lot of writing already had less than zero value to the reader, even decades ago. So that is not their conclusion.

They are really only saying that problem suddenly got MUCH worse.

You said it perfectly - “All writing now is noise”

I was a little cranky when I replied. What I meant was that this flood of LLM-noise already started around GPT-4.5 for many people. I was surprised to read about this now. But it’s good to have a discussion about it at large, anyway.

The other big wallet hardware maker, Trezor, already responded they will majorly cut back their authorized suppliers network.

They want to be certain they have the resources to properly personally verify the operations and trustworthiness of every authorized distributor that they’re vouching for.

https://xxcancel.com/Dantoshi/status/2110000117856248141

These recent teardowns of compromised Ledger wallets are frightening. Because the infiltration of hardware looks very pro. Mass production and standardized parts added.

Even weight difference might not be obvious. He compared weights for one Ledger product but the small variations in weight among genuine devices made comparisons not clear.

Possibly only X-ray would uncover the fake without cracking the case?

https://youtube.com/@MagicalTux

(Warning - MagicalTux is Tibanne or Mark Karpeles. He has a “complicated” legal history, but this looks legit)

This story is a case of someone tampering with devices in the supply chain, before the user first receives them.

What you’re describing is a different thing. It’s an “evil maid” attack, when you tamper with a wallet that is already set up by the user. Also, it would be ultra difficult and expensive.

What you’re saying is possible. But:

It would probably be cheaper to build a total fake Ledger Wallet, then just replace the new real wallet and deliver a fake with a weak entropy generator. Any case, that’s not what has happened this time.

The legit wallets have an online way to verify they are genuine by cryptographic means. A totally new fake couldn't do that.

Thanks. I believe you’re right a totally fake copy couldn’t verify. So that idea was wrong.

(The MITM supply chain attack which has happened succeeds because it preserves the original secure hardware but intercepts the secret data at the display interface. But a low-effort mimic of the entire device like I suggested? That wouldn’t work).

Ledger provides a 'Genuine Check' mechanism to determine the authenticity of the device; however, it explicitly warns that it cannot detect added implants when the device's original Secure Element remains intact.

And as this workaround leaves the original Secure Element intact, good luck detecting the supply chain attack.

I think that’s where the epoxy from earlier in this comment chain is supposed to come in.

Epoxy is useless, we have materials to dissolve cured epoxies for rework in aerospace.

Man can make it, man can break it.

But it's specifically a potting epoxy designed to cause damage to the electronics if removed. And any destruction of the epoxy has to not significantly affect the outward appearance of the device. At worst it's defense-in-depth.