sahmeepee

Playwright has screenshot comparison built in, including screenshotting a single element, blanking specific elements, and comparing the textual aspects of elements without a visual comparison. You can even apply a specific stylesheet for comparisons.

Everything I can see in this demo can be done with Playwright on its own or with some very basic infrastructure e.g. from Azure to run the tests (automations). I can't see what it is adding. Is it doing some bot-detection countermeasures?

Checking if the page behaviour has changed is pretty easy in Playwright because its primary purpose is testing, so just write some tests to assert the behaviour you expect before you use it.

We use Playwright to both automate and scrape the site of a public organisation we are obliged to use, as another public body. They do have some bot detection because we get an email when we run the scripts, asking us to confirm our account hasn't been compromised, but so far we have not been blocked. If they ever do block us we will need to hire someone to do manual data entry, but the automation has already paid for itself many times over in a couple of years.

Not hard drive related AFAIK, but I used to work in a media lab with about 30 mac desktops of around the powermac g4 vintage.

We noticed that the macs were rebooting unexpectedly when certain people were in the room. After a bit of observation we worked out that the call button on our walkie talkies could trigger a reboot from a couple of feet away, which turned out to be an awesome superpower if you'd had a gobful from an especially obnoxious student.

My immediate thought at the start of this article was not DoS but more about harming the company using the chatbot (Company A) by increasing their chatbot bills. In many (most?) cases they will not be hosting their chatbot and will instead be getting it from a 3rd party provider (Company B) who may not even be truly hosting it either.

If the pricing structure is per conversation or per month it would harm Company B, but not the likely target, Company A. If it is paid per interaction it would harm Company A and benefit Company B who just get more paid work.

It feels a bit like cases of rivals clicking on each other's ads to cost them on ad spend, but presumably much lower value than ads.

You would think it would be easy to stop a conversation at n interactions via some other means than relying on the LLM itself, but then you also have to figure out how to stop the attacker just starting more conversations (or passing the output of one of your chatbot instances into the input of another)

I was looking for a way to handle this some time ago, and found that the good people at NIST had already thought about it a lot more.

They made a tool called ACTS which allows you to specify some variables which affect your testing, then generate a bunch of tests with combinations of just n variables at a time where typically n is 2 or 3 but can be higher, perhaps if you're automating the tests or have exceptional patience.

They have some empirical data on how the chance of finding additional bugs decreases as you increase n.

More info here or by searching for NIST ACTS:

https://csrc.nist.rip/groups/SNS/acts/documents/comparison-r...

Lego promote the idea of less too. They make durable products and through Lego Replay they will find a good home for your unwanted Lego bricks.

You wouldn't have them bunched up if you were celebrating the anniversaries, unless all the music you were interested in was from the same year (or only years where 11 September was a Tuesday, say).

Because the OS data doesn't provide addresses, just locations of the postcodes in coordinate terms, so you can't provide the typical website address lookup.

It's mostly used to get clean address data on form inputs. The user thinks they are being given an easier way to enter their address, but really it's so that the PAF can be used to ensure the address lines are used consistently and there are no typos or other rubbish. If the person has to pick their address from a list after entering the post code it's also a data input check of sorts.

I've been a consumer of the PAF for many years through work.

In the bad old days (10 years ago), when we got it from a well known credit check company, it came on CDs. We then moved to "automatic" updates via their rubbish, buggy software to our on-premise install. That wasn't such a huge upgrade, because as soon as you looked away from that software it would stop working.

Now we pay a different, altogether better company to access it via a web service. I would expect that the speed of updates through to websites is now mostly as good as the frequency of PAF releases. Unless this has changed recently, the PAF isn't created by Royal Mail itself, it's outsourced to a company called AFD based on the Isle of Man.

My only complaint with the service we get now is that the value is quite poor - our total cost is somewhere around 30p per lookup on average for addresses which I suspect are rarely even used in our line of work.

Given that OSM has a lot of building addresses and locations (at least it has building names and numbers and the streets they are on, plus their outlines) I wonder how much of the PAF could be recreated from the data in CodePoint Open combined with OSM. If adding more house numbers etc to OSM would help improve that then it could be crowdsourced quite quickly via Street-Complete or similar.

For the purposes of the quick address lookup it doesn't even need to be 100% accurate. If a property is incorrectly listed against two postcodes it won't really cause problems, so the lookup could err on the generous side

It's far from universal and much less used than postcodes for general purpose location, the company just pushed very hard with UK advertising and did a deal with the ambulance service because "everyone trusts the NHS"

...meanwhile, Playwright is sipping piña coladas?

It seems like the choice of Selenium might be a bit limiting if it doesn't allow scrolling, doesn't give access to http status codes and doesn't do full page screenshots or screenshots of specific elements.

If you're invested in Selenium for other reasons I can see why it might not be worth switching. I can't test it right now but I'd be interested to see what a simple Playwright test does with this: just scroll to the bottom and back to the top, wait for the lazy elements to load, then take a full-page screenshot.

Those 16k screens would need to be very large and very close to the viewer for the human eye to be able to resolve the individual pixels.

You really don't need mapping to get reasonably efficient cleaning. We have a robot that's nearing 15 years old with no mapping, but the algorithm it uses is pretty effective. The mapping represents huge over-engineering for what is a pretty simple problem.

This is similar to the idea that if you scaled the Earth to the size of a standard size 5 football and dried it off, you would barely be able to feel the mountains or trenches on the surface. The water is therefore a very thin film over the land in those terms.

I was just relieved when it wasn't vulva

So I believe this rules out a supply chain attack where someone buys a bunch of CPUs, infects them, repackages them and sells them as new.

If you can't do that, then this feels significantly less problematic.

I think when you dig into people's self-promotion pitches you often find they're less impressive than they initially appear, so taking them at face value might be bad for your morale.

I have no problem with a 22 year old blowing their own trumpet to get along in the world, but as an example from that bio, the AI generated news site with 4m views is effectively just a load of SEO spam so is going to get a lot of views in 5 years. It has 2 linked articles about it. I can't read one, but the other is completely trashing it both as a concept and in its execution. If I'd made it at 22, I'd probably be pleased enough to put it on my bio of course.

I don't know if they still have it, but the AI playlist builder on Spotify could do stuff like that. If you play a regular playlist on Spotify called Reggae or Rare Reggae or obscure Reggae you can bet your life it has a few Bob Marley tracks on it. The AI thingummy was able to restrict to less popular artists only.

I agree strongly. Listen to the radio stream for any reasonably niche artist (especially if they have had a single "hit") and you will be treated to 3 or 4 different mixes of the same song within the first hour. For some purposes Spotify is just unlistenable.

"Other things" including intentional commercial biases presumably?

No matter what I do in Spotify, under several different rounds of accounts, it always seems to gravitate towards the tastes of the general public, i.e. some form of mass-market pop.

Their recent "ai" assistant was a slight improvement because you can ask it for less popular music which is typically better for music discovery.

Importantly for music discovery, Tidal make it relatively easy to browse by record label so if you're not listening to major label pop stuff you can pretty easily find a bunch of artists in a similar niche. If that is possible on Spotify they've made it hard enough to access that it might as well not have been there.

Last.fm was good, but inevitably went downhill when it was bought out. As they had to commercialise more and more from the early days as a uni project that became audioscrobbler, with a tiny userbase, they followed the standard Doctorow model of capitalist decline. They had a sweet spot a few years in when they had plenty of data flowing and they hadn't yet messed up their APIs. Right now I use it to log my listening but I'm waiting for the email that says I'll need to move my data elsewhere.

I remember going to an event in 2006 called "Mapchester" in Manchester, UK to gather data for OSM with GPS units and digital cameras. It's come an incredibly long way since then!

This will not resolve your second issue as it replaces timbre but not accent.

Adding to the list of evil applications with multiple licensing tiers that force you to pay for the top tier to get SSO.