Honest sellers pay VAT, and scofflaws get sales. Yeah they gotta throw down the gauntlet here or else VAT is only for suckers.

We recently went to Niagara falls on the Canadian side and it was fun. Canadian sales taxes and fees took some of the currency difference, but yes we had a decent deal on a steak dinner in the tourist trap.

Yes plausible text prediction is exactly what it is. However, I wonder if the author included benchmarking in their prompt. It's not exactly fair to keep hidden requirements.

Attributing these to "hidden requirements" is a slippery slope.

My own experience using Claude Code and similar tools tells me that "hidden requirements" could include:

* Make sure DESIGN.md is up to date

* Write/update tests after changing source, and make sure they pass

* Add integration test, not only unit tests that mock everything

* Don't refactor code that is unrelated to the current task

...

These are not even project/language specific instructions. They are usually considered common sense/good practice in software engineering, yet I sometimes had to almost beg coding agents to follow them. (You want to know how many times I have to emphasize don't use "any" in a TypeScript codebase?)

People should just admit it's a limitation of these coding tools, and we can still have a meaningful discussion.

The training data is full of ‘any’ so you will keep getting ‘any’ because that is the code the models have seen.

An interesting example of the training data overriding the context.

Then you add a biome rule to say "no any ever" and the LLM will fix it before claiming the job is done.

Yeah I agree generally that the most banal things must be specified, but I do think that a single sentence in the prompt "Performance should be equivalent" would likely have yielded better results.

I took a taxi ride from Niagara (ON) to Buffalo. The Canadian driver really was leery of Americans and I apologized for everything. It's a dang shame, and I don't blame you all for feeling this way.

Bit off-topic, but how easy was this to do? We need to do the same crossing to pick up a rental car from Buffalo.

I couldn't arrange it via app, so it seemed impossible at first. However, I asked the bellman at the hotel, and he called his taxi driver friend. I kinda overpaid from what I can tell, $100 american, but he just drove us across the bridge, passports were checked super quickly by the American side, and we continued on to Buffalo in about 40 minutes total.

Thanks! I'd seen Uber etc. won't cross the bridge, so looks like talking to an actual human is required :)

Do you Americans realize it means absolutely nothing to us when one of you "apologises for" Americans? You do that for you, not us. It's weird and gross. You don't speak for Americans. Americans speak for Americans, and the message is loud and clear.

Well in the taxi or seemed to be the right thing to do.

Payment processing networks are not free to build or operate. There are necessarily fraud controls and transaction reversals that require human oversight. This all costs. Nations can and should build this infrastructure, but in the absence, a payment processor is going to charge interchange. Otherwise why would they bother.

Uninvolved companies wouldn’t bother, but outlaw such fees and lenders would still build this kind of infrastructure so they could make money charging people interest.

The commentariat senses blood in the water and will criticize Ars Technica no matter how they respond here. It seems fine. The author really paid the price. I trust Ars to be extra vigilant to this going forward.

The thing that needs to be age banned, or really just banned, is algorithmic feeds with infinite scroll. Kids (and adults) need to just interact with their friends, and block all the bait.

For adults, I think a legal opt-in-only policy would work well. And require reconsent with every major algorithm change.

It's even sillier after looking into it. Of the 4 people listed that died the same date as the insurrection attempt, 1 was shot (already mentioned), 1 died of overdosing on meth, and the other two both were over 50 and had heart attacks. Not to say being exceptionally out-of-shape or meth-addled has zero demographic connection to the riot, but...

s/Donage Page/Donate Page/g

This is my experience of it too. Perhaps if it was chunking through a large task like upgrading all of our repos to the latest engine supported by our cloud provider, I could leave it overnight. Even then it would just result in a large daylight backlog of "not quite right" to review and redo.

I think that's the issue I have with using these tools so far (definitely professionally, but even in pet projects for embedded systems). The mental load of having to go back through and make sure all of the lines of code do what the agent claims they do, even with tests, is significantly more than it would take to learn the implementation myself.

I can see the utility in creating very simple web-based tools where there's a monstrous wealth of public resources to build a model off of, but even the most recent models provided by Anthro, OpenAI, or MSFT seem prone to not quite perfection. And every time I find an error I'm left wondering what other bugs I'm not catching.

What I tell my kids is: You know how when you ask AI about something you know very well, how its answers are always somewhat wrong? It's like that for things you do not know very well too.