mathijs

I've used simpler agents like Copilot and Devin/Windsurf/Cascade/whateveritiscallednow, mainly in IntelliJ, and depending on the model, they starts showing behaviour that is at least remotely like this.

Example: put the agent in Ask mode (so it can't edit files) and you'll see it try to edit files anyway. The train of thought shows "something went wrong editing the file, let me try a different way" and it'll start spewing out bash files or Python scripts that try to edit a file. None of it works or can be executed, but still.

Cheaper models often ignore the available function calls to find and edit files in the IDE, and will start asking for permission to execute grep and sed commands, as well as trying to echo entire bash or Python scripts to file again.

It is not exactly like an agent autonomously trying to hack Huggingface, but it is a way of frantically looking for a solution because 'giving up' is not what LLMs are trained for.

Hey, this is my blog, thanks for sharing!

I've been finding myself increasingly annoyed by non-coding managers thinking AI is going to bring 190% reduction of cost, or will replace entire divisions of coders. It's a helpful tool, sometimes, but sometimes also a complete and utter tool. So I wrote this rant about it.

I've used it for 15 years and ST is still the first thing I install on a new PC.

All thoughts, meeting notes, journals, blog post drafts... everything is jotted down in ST first. I even went as far as writing my own to-do list syntax highlighter[1] which is the main reason ST is always open, at home and at work, even though I mostly use VSCode and IntelliJ for coding nowadays.

[1]: https://github.com/mlagerberg/todo

Exactly. A video on YouTube posted '1 month ago' is not the latest episode I'm looking for. Much quicker to interpret than a timestamp. Sometimes a fuzzy date is enough.

Especially being from Europe: it is never clear if 10/04/23 means October 4th or April 11th unless you know which date format is being used.

Android Studio offers this out of the box since the beginning of this year, I believe

The apps do, but the website does not. music.google.com used to have native support, but to cast music.youtube.com you have to cast the tab, which is far less than perfect. (Instead of streaming directly from the CDN to the chromecast, casting a tab will stream from your desktop to the chromecast. It works so poorly that sometimes even the pitch changes slightly, like an old timey recording.)

And still no Chromecast support.

Makes sense. The principle of 2FA is to combine 'something you know' (a password) with 'something you own' (your phone). I guess the biometric lock is 'something you are' on top of that.

I agree with this comment completely. Adding a biometric lock would turn it into 3FA.

Not sure if HN allows to plug your own apps, so please forgive: I made an app a while ago that aims to replace Google Authenticator for some of the reasons mentioned: it allows to back-up and transfer tokens without creating a large attack factor. Not having sync is a feature in this case as well. In fact, the app does not even have the internet permission enabled, so it utterly unable to phone home. Transferring backups does require a biometric lock.

It is also entirely free, so I'm only posting this out of pride of my own work: https://play.google.com/store/apps/details?id=com.pixplicity...

That's a plastic 'Flippo'. Similar to cardboard milk caps but they came in bags of potato chips and are made of plastic.

Not necessarily dark though... as an app developer I'd actually really want to hear the feedback from the people who dislike my app. As in any business, oftentimes the most disgruntled customer is the one that can provide the most useful feedback.

An even smarter way (in my opinion) of getting 5 star reviews is by showing the dialog only after x hours of use. Users most likely to rate badly will have the app uninstalled before the dialog shows.

I've used Xamarin for several projects and I greatly regret that decision. Granted, most of my hatred is targeted towards Xamarin Studio and Xamarin.Forms so if you use Visual Studio and build only native UIs then maybe your experience will be better.

However, in my experience, both Xamarin Studio and the build system are buggy as hell. Random or inexplicable build errors, things that break during updates, incompatibilities with official Android support libraries... I find myself doing 'clean project and rebuild' to fix random errors, or switching between alpha, beta and 'stable' channels all the time depending on which one does not have the bugs that I'm running in to.

Xamarin.Forms is simply a disaster. Because it aims to unify the apis for the UIs for various platforms it boils down to only the most common denominator of those platforms. And then makes it worse. Not only is it buggy, it is also very slow and incredibly limited. In our office we're keeping a list of all of Xamarin's silliness we encounter, here is just one of those:

    "Clicking a Button changes its text alignment from center to left-aligned; it requires writing a custom Renderer to solve this."

I admit that Microsoft is usually quick to fix those bugs, but it doesn't instill much trust in the system if you're constantly running into issues. Many days I am literally working 50% of the time on my app and 50% working/fighting my way around Xamarins issues.

I'd love to hear from someone using Visual Studio if their experience is more positive, but my advise is: please stay away from Xamarin.Forms and Xamarin Studio as much as possible.

Thanks! I didn't know that, but I'll check it out. Thanks!

Hi, and thank you for Wallabag! I've been happily using the self-hosted version for quite a while now. At some point I forked[1] the Android app but development stalled when I needed an API to be able to improve the app further. So this v2 is great news!

Just an FYI: in the blog post you mention that the login for the preview of v2 is wallabag/wallabag, however I get 'bad credentials' when trying that.

This new Material Design version looks awesome! Eagerly looking forward to v2 becoming stable so I can upgrade.

[1]: https://github.com/monkeyinmysoup/wallabag-android

Ah, thanks, that makes sense.

I think what they mean is that you don't have to type any javascript yourself, such as setting onclick-listeners. At the top of the page it reads "CSS powered (jQuery fallback)", however, if jQuery fails to load (e.g. when visiting the https-version of the same page), it doesn't work.

What I also don't understand is how this plugin was in any way 'missing'. It seems we had plenty already[1].

[1]: http://www.designrazzi.net/2014/free-jquery-accordion-menus-...

We prefer native 99% of the time. Performance of HTML is easily outmatched by native on both iOS and Android, and the multi-platformness of HTML is usually limited if you want to adhere to UX guidelines and conventions that are platform specific.

Perhaps they're just keeping an eye on their competitor?

Ironically enough, I went to log in to HN specifically to upvote comments in this thread, forgot my password and had to use the 'email yourself a new password' function.

Granted, it sends a randomly generated password. And granted: it's probably meant for one-time use only (I presume). However, logging in with the new password does not automatically prompt (or even force) to change it to something of my own choosing. Result: hit 'remember password' in Chrome and forget about it. Now my active HN password is in my inbox, in plain text.

When I was young my dad's computer had Office with a localized version of VBA, in Dutch. If I'm correct it was Office 95. I've tried to use it for a while but it was horrible; even as a non-native English speaker I couldn't get used to the translated statements and kept mistakenly typing their English counterparts.

Here's what it looked like. If regular Visual Basic looked like this:

IF x == 1 THEN '... ELSE '... END IF

the Dutch version would be: INDIEN x == 1 DAN '... ANDERS '... EINDE INDIEN

Odd thing that 'indien' is a rather formal and slightly archaic version of the (much shorter) equivalent 'als'. I don't even remember how FOR, WHILE, and all the other statements were translated.

I like the reasoning behind this and I really welcome the simplicity of giving feedback, but the way it is implemented here does not really make things clearer. The link makes me think I'll get a simplified summary of the paragraph, instantly. It's not clear that it sends an email to the writer.

Also, the writer will never know if a feedback-email comes from a genuinely confused reader or from someone like me who simply though 'what does this button do?'.

A solution would be to show a message after clicking the link, saying something like 'thank you for your feedback, we've send an email to the writer' together with a Cancel button.

Coincidentally I've asked a question about login system just two days ago (http://news.ycombinator.com/item?id=4225270) where I mentioned BrowserID as one of the options. Apparently I hadn't properly checked out BrowserID/Persona because I thought of it as an OpenID replacement, not an email+password login replacement.

I'm positively surprised to see that you found a 50-50 balance between BrowserID and Facebook; I expected a lower pickup rate on BrowserID. So, I guess I've decided what login system to use for my project(s).

Thanks for the information.

Thanks! I didn't think about LastPass, but since the website will be targeted (mostly) towards a hacker crowd it makes sense that many of the users will use a password manager like LastPass. I'm not a LastPass user myself but I guess that takes away a bit of the hurdle from a user's point of view.

This is not a website with a searchable archive, but it collects websites that send you an email saying 'you just signed up and this is your password': http://plaintextoffenders.com/