I visited Vietnam a couple of weeks ago and they have a home-grown (probably heavily government-subsidised) Vinfast company, which specialises in EVs. I'd say at least 50% of the cars in the cities there are VFast EVs now (though there are still ten scooters for every car, and not many of those are EV). They're perfectly adequate cars; equivalent to BYD or similar.

Despite being something of an EV enthusiast, I'd never even heard of them before.

They're starting to try to export them (they even manufacture them in North Carolina now); it's something to keep an eye out for. Vietnam's a rapidly developing economy, basically following in the footsteps of China, so it's a good time for them to ramp up their auto industry.

Visa screening processes (and indeed most screening processes for anything anywhere) are flawed fromt the get-go.

Most (99%+) people are completely harmless and won't cause any problems.

Most broad demographics (eg: Palestinian, Christian, Australian, Gen-X...) will also have 99% completely harmless people.

Trying to identify trouble-makers only works if you have non-demographic information about the person you're identifying... but most (90%) people won't have a public-enough life for that information to be available.

Only ~73% source of people in the world have uses the internet, and of those, ~90% never contribute more than the occasional comment. So analysing social media use won't help much.

For crime, estimates vary but in first-world countries roughly ~5-30% of people have some kind of criminal record, so maybe that helps; but most of those crimes would be the sort of trivial stuff that really shouldn't prevent a visa (eg: traffic offences).

So screening will only pick up the really obvious stuff (bad criminal records, news-worthy offences, certain types of celebrities). It won't pick up the quiet sociopath or the devious white-collar grifter or the desperately-in-debt-to-criminal-gangs.

And the expense and the delays and the false positives are just incredible. I don't think we get good value-for-money for it, or good value-for-delay, or good value-for-inconvenience.

No, in fact it's sort of the opposite - the social media companies have to prevent kids from logging in, but they don't have to prevent them accessing it. So, if anything, there will be more no-login access to sites because the SM companies still want kids' eyeballs (and just because they're not logged in doesn't mean they're anonymous; Facebook will still know from tracking cookies, etc).

I basically disagree with the legislation; it's ineffective and poorly targetted. Kids can bypass the blocks in many sorts of ways, and the loss of anonymous accounts is actually really bad; I've had a spare throwaway Facebook account with a fake identity for ~10 years, and it could well be caught by this. Also, how can a kid learn to be safe online if they're not allowed to use half of the web? And how can parents use parental controls to restrict a kid to safe parts of Youtube if the kid can't be logged in to it? And of course 4chan isn't banned.

That said, it's a social discouragement: "I can't let you use Snapchat, it's illegal" is a good excuse for a parent to prevent Snapchat usage if they catch the kid using it, while "but all my friends use it" becomes a less effective comeback because all of their friends' parents have the same excuse. It's not going to stop a determined kid of an inattentive parent, but it is a social pressure that might help to reverse some of the crippling SM addictions that way too many kids (and adults, tbf) are hit by.

It's not an uncommon approach, but the 17 mins (33%) varies. My physics prof went for 50%, one particularly picky maths prof I knew went with 80%; you really had to know the material for his exams.

And not only that, but road surfaces are designed with expected traffic patterns in mind - stronger surfaces for highways, weaker/cheaper for side streets.

You can choose your level of detail for these sorts of calculations but you'll always make some assumptions.

Our university uses https://github.com/dropbox/zxc... , which does that as well as some simple algorithmic transformations (eg: pa$$w0rd). We don't say, "it can't be stupid" but we do say "that password is considered insecure".

That said, MFA almost makes passwords obsolete; the threat landscape for guessing a password online is far wider than the landscape for having access to the student's unlocked phone.

And yes, we support Passkey these days too.

A bit of Googling (which you could have done, were you so inclined) suggests Barcelona and Majorca.

And before you say "they're not deserts yet", he did say "forming apace", not "currently here" so don't strawman him (but hey, maybe I'm strawmanning you). For evidence of "forming apace" you can look at the data in the study linked from the above article - already measured decreased overall rainfall, and increased droughts on top of that (ie: all the rain is concentrated in one time period, aka floods).

Of course, he also mentioned a string of other climate issues, none of which you argued against, so you must agree with him on those. So I sure hope you're as alarmed by this as I am. ...I get sick of the "show me the evidence" crowd. There's plenty of evidence and no matter how much I show them they don't care, because they're not actually ready to listen.

If you think you are ready to listen, but you still don't believe climate change is a threat to humanity's future, consider this question: what realistic evidence would convince you? If you don't think such evidence could realistically exist... then you're not ready to listen.

Climate change doesn't look like "Day after Tomorrow". It's not going to be 10 degrees hotter today than yesterday. The effects are large-scale, long-term, and subtle, and will slowly destroy human civilisation. The more we do now to slow it down, the better it will be for everyone in the future.

Right, but it may well be just as easy to find a place to rent that's $1800 (maybe further from the supermarket/restaurants) and keep the Doordash.

You can make any small value into a big value by accumulating it over a year, but it's still a relatively small value compared to your other accumulating costs.

We're financially secure (own our house outright) and we get doordash simply because we don't have the time or energy or interest to clean or wash up. We do feel a little guilty about it (the business model does seem pretty unethical), but our guilt is not because we could be spending the money more wisely.

They'll remember after the first couple of times.

Mobility scooters are much heavier. They're also less manouverable and have slightly restricted sight lines compared to your typical walking person. The 4mph reduces your chance of running over babies crawling in front of you.

I have a friend who suffered sever spinal injuries in her 20s because a mountain biker crashed into her when she was walking down a hill. The rider had lost control (they didn't mean to be going that fast) but he got away with minor grazes, and she's the one who can barely walk.

Even without electric assistance, the mass and speed of an out-of-control bicycle + rider is not to be underestimated.

It's short-sighted. How expensive are people's lives going to be when the earth is 3 degrees warmer? That's not "turn up your air-conditioners", that's "massive natural disasters every year" and "farm yields drastically decline" and "some countries become unlivable" and "fish stocks collapse" and "water wars".

If you want to forget about the ethics of making many species extinct, we can just play the economics game instead - but it still doesn't look good.

There's Nepenthes, and it's open source, though it sends back slow, Markov-chain nonsense rather than actual falsehoods.

Plot twist: the attacker already got in, and that's why they pushed for cloud-saving to be the default.

Yes, but my point is that populating the database would be a front-end generated by an LLM, not (or not primarily) end users directly chatting with an LLM.

You'd still have a form, with validation rules, but the form would have been made by CRM admin asking an LLM to generate it, subject to validation rules in the system, then signed off by HR and Legal, then deployed for people using it. The form might have a chat option where you can ask the LLM to help you fill in the form if you wanted to, though, and because the LLM has a model of the underlying data it can answer those questions accurately.

I suspect we're still a way off from all of this working smoothly though, and I'm not really sure it's possible to solve the privacy, security and data lifecycling issues reliably (though we have those problems with existing ERPs all the time too).

I think the idea is you populate your vector database with facts, and use the LLM to process queries against it. Some of those facts will be information about customers, some of those facts will be policies/laws, some of those facts will be point-in-time data. Then you can ask your CRM AI: write a custom query to gather all policies/laws that relate to customer X in situation Y. Save that query, then present it as webpage Z. Do that for everything anyone asks a CRM, and you have a bespoke CRM, which you can update with new features whenever you want.

The policies would have to be super-reliably encoded. Maybe you'd have red-team AIs that try to poke holes in the query. You'd certainly have (non-LLM) models that exist solely to analyse the graph to find what else needs adding or removing or improving. But at the end of the day the actual solution is a chain of logic that can be followed by hand if need be, not a hallucinated string of words.

Think of it as a graph database with a natural-language query front-end; you have your data repository, but it's in a complex and changeable form that can capture more of the nuance and be expanded more naturally than your average SQL DB.

No idea if this would actually work but I wouldn't dismiss it entirely.