sahmeepee

OneNote has some predictive text gubbins now.

Umm, not sure which version of notepad.exe keeps a lock on open files. None that I've used in the last 25 years.

In some colleges it's the bar, e.g. https://www.seh.ox.ac.uk/discover/conferences/dining-at-st-e...

"Finally, we acknowledge the correlational nature of our findings. Sleep regularity may be both a cause and marker of premature mortality risk."

The source data is from the UK so it has nothing to do with the health insurance attached to specific jobs. Healthcare is covered via general taxation in the UK.

A lot of those date and time pickers will fail WCAG testing as well.

In most cases, GDS now use just 3 text fields for day, month and year, but recommend pickers for dates close to today's date e.g. for booking an appointment, because a picker is easier if you don't know the exact date without reference to a calendar.

Sadly they don't currently have a recommended picker, but there's a useful discussion of what has been tried here:

https://github.com/alphagov/govuk-design-system/discussions/...

Amazon's product search is largely useless, at least in the UK, now. It typically returns many items that don't include your search term, are outside the price range you specified and are not sorted in the order you asked for. The top half the list being "sponsored" items is just the enshittified cherry on the turd cake.

Probably not the first AI wrapper around Playwright this week, and certainly not the first this month.

I think this use case of automation in a BPA sense is more compelling than using it for test automation, because the latter is much more concerned with the precision and repeatability of the process. For the BPA task, arguably you care only about the outcome and it often doesn't matter if it gets there via some crazy route.

Part of the problem for me is that your example video shows a big wodge of prompt that had to be written to make this work and then a few kb of payload data (parameters) in a plaintext, non-csv format. If the expectation is that this replaces someone just using Playwright with codegen due to that being too technical, I'm not convinced there is a huge group of people who can manage one task but not the other.

Furthermore, you are expecting them to pass over their website login credentials and apparently their credit card details too, in plain text. You had better have a very solid idea of how to handle that sensitive data to avoid serious consequences if your users' skyvern accounts are compromised.

I think the frequency of website redesigns is oversold by people producing these LLM-driven Playwright wrappers, especially when targeting old-fashioned or government sites. As an example, we have had a suite of lengthy Playwright browser automations to interact with a government site for a few years and have had to maintain them only once, when the agency's business process changed. The prompt would also have needed to change had we used Skyvern, as would the payload, because the process was different. The difference with the Playwright automation, though, is that we could use assertions to verify steps had succeeded/failed and data had been recorded correctly, so we would know the process needed updating. I can't see that option in Skyvern which would have me worrying that process changes would be overlooked and we would unknowingly start entering the wrong data or missing steps.

I think the Street Complete approach is much more practical than anything voice based: slightly gamified "quest" based entry which lets you tackle an area or a data type in detail.

It's pretty much impossible to do any meaningful and accurate data entry while riding a bike, whether the interface is touch, voice or whatever. You need to be able to verify that the phone has accurately captured what you were trying to tell it, not just take it on faith that it's worked, and that means looking at the screen carefully. Good luck to them making this work reliably.

Thanks for the detailed response.

The Playwright dev team would probably say that you should avoid using IDs as selectors and instead favour the use of selectors based on user-visible aspects, e.g. "a link including the text 'cat'" or "a button with the label 'register now'". That way your tests are immune to under-the-hood changes the user is oblivious to. The range of selector options (locators in their world) are a real strength of Playwright.

Thanks for this. In general I'm in favour of products removing the dates so that people don't chuck out perfectly good food based on superstition/misguided obedience, but an annoying side-effect is that if we have two unopened packs of the same product it's hard to know which is older. By sheer bad luck, an old pack could hang around indefinitely and go off.

I was thinking a possible solution would be a simple numerical counter (e.g. days since 1 Jan 2000), which can't be easily decoded by eye, but allowed you to rank items easily by age.

The axe element will be the same each time, but you can do that without any AI shenanigans - just run it in a normal Playwright test.

My question was really about the page interactions and the assertions being driven by AI: if they are going to be generating different code every time the test runs, how can you have any confidence in the test not having false positives and false negatives at least some of the time, unless you read the generated script each time?

That sounds like a lot more work than just writing the test once in the traditional way (codegen or manually) and tweaking it only when there's a breaking change to the page.

If people are genuinely using this approach then there must be something I'm missing.

Playwright (and axe) is a good option, but how do I have confidence it's performing the test correctly and repeatably? If the test seems flaky how do I know it's the software and not randomness of the AI part? I want tests to be predictable.

Thanks for this - it's a very accessible format.

I stumbled on this fantastic piece promoting the use of UTC time, but incidentally giving a great history of the standardisation of time more generally and the adoption of 24h clock notation over AM/PM.

https://en.wikisource.org/wiki/Popular_Science_Monthly/Volum...

In this country many teachers have been taught that it's important for kids to be able to use a range of approaches to solve a problem.

They then force them to use each of the prescribed set of methods even when they are totally inappropriate for the task. Any deviation from the method they are told to use for that question is wrong. No creation of your own adapted methods is acceptable.

In other words, the teachers were taught that kids should use different methods, but seemingly weren't taught why.

I thought that was the purpose of adding noise (in the mitigations).

You might be interested in this app to try it for yourself:

https://github.com/nisargnp/DeadReckoning

I read a paper some time ago which suggested that mobile devices are still nowhere near good enough to use this e.g. for navigation around a shopping mall.

If 30% is climate levy then that's some rather cheap electricity, because the climate levy is less than 1p/kWh.

https://www.gov.uk/guidance/climate-change-levy-rates

This is what happens when people treat the Telegraph as news rather than a source of disinformation used by wealthy vested interests to manipulate public opinion.

I'm referring to the typo in your very first heading, unless datat visualisations is a term I don't know.

Datat visualizations?

Great idea for a game and very playable. As others have said, it is much more interesting than 2048 due to the reduced repetition. It still seems beneficial to stack high value pieces out of the way (e.g. your biggest in the corner and then next biggest on an adjacent edge square). Towards the end it was getting quite frustrating having rooks appear in corners, but a good kind of frustrating!

Could the score be visible during the game to give a sense of the scoring rules? I didn't know the king eats pieces until I read the comments here... I would expect that to lose you points, but hard to tell unless the scoring is maybe documented somewhere I missed?

Apparently he is worth 1024 so you can take him when you have a 1024 piece, and win. I didn't realise this (or that he takes pieces when cornered - I avoided that, assuming a stalemate would mean losing) so played it right through to a regular 2048 which took a verrry long while.

As the article says, you don't need an active subscription to be vulnerable. In this case it seems that if the model supports the features at all, you are vulnerable.

This makes sense, because they want people to be able to subscribe to their services later without having to visit the dealership, so they make it possible to remotely enable the service.

I'm not sure if you can buy a tinfoil hat for a car.

I'm not sure that Luddites really represent fighting against a process that's flawed, as much as fighting against one that's too effective.

They had very rational reasons for trying to slow the introduction of a technology that was, during a period of economic downturn, destroying a source of income for huge swathes of working class people, leaving many of them in abject poverty. The beneficiaries of the technological change were primarily the holders of capital, with society at large getting some small benefit from cheaper textiles and the working classes experiencing a net loss.

If the impact of LLMs reaches a similar scale relative to today's economy, then it would be reasonable to expect to see similar patterns - unrest from those who find themselves unable to eat during the transition to the new technology, but them ultimately losing the battle and more profit flowing towards those holding the capital.

Aaah, but can it show you into a room with a toilet, sink, shower and bath in it and tell you "This is the bathroom"?

If not, you'll never unseat the seasoned professionals of the UK real estate sector ;)

It actually varies a bit in the UK (always for the seller as you say) with a possibly surprising relationship with property value: properties at the top end typically pay a higher % than those in the middle or at the bottom.

> These may seem like marginal differences

Yep. A difference of a couple of percentage points even with a sample size of 20,000 is tiny. While they have taken it to mean that hybrid is "better" than in-office, I would take it to mean that they are essentially identical across the breadth of the survey.

For your particular industry or organisation there may be significant variation, but you'll need to find that out for yourself.

Or check out "money printer" on github: a tongue in cheek mashup of various tools to take a keyword as input and produce a youtube video with subtitles and narration as output.

Prior to Google we had Altavista and in those days it was incredibly common to find keywords spammed hundreds of times in white text on a white background in the footer of a page. SEO spam is not new, it's just different.

Sounds like a prooblem that can be solved with a Playwright script with a bit of error checking in it.

I think this needs more elaboration on what the Finic wrapper is adding to stock Playwright that can't just be achieved through more effective use of stock Playwright.