I agree that the locking down is truly stupid. For what it’s worth, the reasoning for locking down mobile apps is allegedly that mobile users are a less technologically competent demographic than desktop users. I do not think so myself, given the difficulty in trying Graphene vs. Desktop Linux.
“Installing alternate OSs” is juicy bait for “tech enthusiasts” who know just enough to be effectively worse off than someone with a browser, yes, and at its core is this holier than thou attitude.
I don't agree that it is stupid. Both banking on a Windows PC or on an unlocked + rooted phone is potentially catastrophic. Windows because of the prevalence of malware, unlocked phones with custom AOSP forks because people download 'ROMs' (as they call them) from the most shady sites.
Once 10,000s of Euros are siphoned from a bank account, it's usually the bank that has to deal with the mess. Especially if they cannot prove the transactions were done in on an insecure platform.
Phones are generally safer (though there is a huge variance between the safety of different Android phones) because they use verified boot and strong application sandboxing.
I think it is possible to believe the following two things a the same time:
- Banking apps should only run on locked phones with secure boot.
- Banking apps should not be limited to the Apple/Google duopoly.
The solution is that there is some validation of alternative OS vendors, e.g. in the form of an audit, and that banks are required to approve apps on their platforms after the audit. This would be fairly straightforward tech-wise, because e.g. GrapheneOS supports remote attestation, but banking apps need to add/allow the hashes of the official boot keys: https://grapheneos.org/articles/attestation-compatibility-gu...
Needing to use a verified boot chain with keys that the bank trusts is essentially the same as using the authenticator device from said bank, except this one costs 100€ or more, has a microphone and camera built in, and you use it for private messages as well. That's not a future I want to live in
We have secure hardware already, it's called a smartcard and is what you find in all bank cards, SIM cards, authenticator devices... my phone is my phone, not a second factor, or at least I (as a hacker/tinkerer) don't want it to be that way, just like with my desktop which is also not the bank's to mandate whatever from
Somehow they got the memo for devices where it is normal to have admin permissions, but for mobile devices the two big tech companies successfully scaremongered non-techies
Needing to use a verified boot chain with keys that the bank trusts is essentially the same as using the authenticator device from said bank,
It's not, because even though the authenticator is secure, you are entering the auth codes in a browser in general purpose desktop OS with (if you use Windows or desktop Linux) little to no sandboxing outside the browser. You are one malware app (or NodeJS package for tech users who claim they'll never download malware) for your session getting hijacked.
The sad reality is that phones (and some tablets) are the only relatively secure computing environments that we have. Thanks to Windows with it decades of piled up legacy and Linux with large sandbox and secure boot-hating parts of its community, we cannot have nice things.
(The part about the Linux community, which I'm also part of is a generalization, but the hostility against Flatpak, secure boot, etc. is pretty big.)
That seems wrong. If malware can fake what the authenticator shows me, the authenticator is broken!
That's not what I am saying. The authenticator is irrelavant to this attack. If your machine is compromised by malware, the malware could take over the browser session, regardless of how you log in.
Phones are better protected against persistent malware because every application is sandboxed (harder to escalate) and much more of the boot chain/OS is validated (harder to persist).
The authenticator preserves account integrity with the compromised host attack you describe. The device I have shows something like "you are authorising a transaction of 1337€ to RU07BANK012345678, y/n?". What an attacker can do is read along while I log in, but not modify data
The server generates the challenge that's sent to the authenticator. The attacker can modify and replace it by being in your browser and show any text on your computer screen, but the authenticator will either show the truth, or the approval code it generates doesn't match the server's challenge
> That's not what I am saying. The authenticator is irrelavant to this attack.
If you need to change the security measures (take out the authenticator) in order to be able to mount an attack, maybe that means the security measure is working? xD
Trying to understand your point here. If you're merely saying that phones have better process isolation then I can only agree, but I wasn't saying that it doesn't. You can use online banking on your phone OS if you like, or use Android on your laptop. The comment you replied to upthread said that I'd like to have ownership of and freedom within my own hardware, in order to have privacy. When banks require that my phone is DRM'd with some keys from Google, Samsung, or Apple, then suddenly that has a lot of consequences for what I can and cannot do with, or inspect about, the device. Using an external authenticator, which they can attest to their heart's content, is the solution that I'm using and aligns with all parties' goals. Banks don't need to require that everyone's phone is locked down in order to use the banking software, just like it isn't in the browser, while still meeting their security goals
Seconding this—I had to wait a little bit to download it and play around and have some fun with it. I didn't mind.
What I appreciate the most about this string of comments (from OP) is that digging into "doing it for fun", hosting on your own machine, wanting simplicity for you as the maintainer and builder. This has been a big focus for me over a number of years, and it leads to things being not efficient, or scalable or even usable by others—but they bring me joy and that is more than enough for most things.
The reality is that there are of course ways to make this more efficient AND it simply doesn't need to be.
Good job on making something that people are clearly interested in, it brought me some joy clicking around and learning some things.
If you want it to be more than just this, of course you'll have to make it faster or have it be a different interface—installable offline typa thing so we can expect a bundle download and be fine with waiting. For example I can see this as a native app being kinda nice.
If you don't want it to be more than this, that's okay too.
Which local AI do you use? I am local-curious, but don’t know which models to try, as people mention them by model name much less than their cloud counterparts.
I'm frequently rotating and experimenting specifically because I don't want to be dependent upon a single model when everything changes week-to-week; focusing on foundations, not processes. Right now, I've got a Ministral 3 14B reasoning model and Qwen3 8B model on my Macbook Pro; I think my RTX 3090 rig uses a slightly larger parameter/less quantized Ministral model by default, and juggles old Gemini/OpenAI "open weights" models as they're released.
I think that this summary is oversimplifying: The rest of the blog post elaborates on how the author and Masley has a completely different interpretation of that bullet point list. The rest of the text is not only examples; it provides elaborations of what thought processes led him to his conclusions. I found the nuancing of the two opposing interpretations, not the conclusion, the most enjoyable part of the post.
(This comment could also be shortened to “that’s oversimplifying”. I think my longer version is both more convincing and enjoyable.)
I feel like your comment is in itself a great analogy for the "beware of using LLMs in human communication" argument. LLMs are in the end statistical models that regress to the mean, so they by design flatten out our communication, much like a reductionist summary does. I care about the nuance that we lose when communicating through "LLM filters", but others dont apparently.
That makes for a tough discussion unfortunately. I see a lot of value lost by having LLMs in email clients, and I dont observe the benefit; LLMs are a net time sink because I have to rewrite its output myself anyway. Proponents seem to not see any value loss, and they do observe an efficiency gain.
I am curious to see how the free market will value LLM communication. Will the lower quality, higher quantity be a net positive for job seekers sending applications or sales teams nursing leads? The way I see it either we end up in a world where eg job matching is almost completely automated, or we find an effective enough AI spam filter and we will be effectively back to square one. I hope it will be the latter, because agents negotiating job positions is bound to create more inequality, with all jobs getting filled by applicants hiring the most expensive agent.
Either way, so much compute and human capital will go wasted.
> Proponents seem to not see any value loss, and they do observe an efficiency gain.
You get to start by dumping your raw unfiltered emotions into the text box and have the AI clean it up for you.
If you're in customer support, and have to deal with dumbasses all day long who are too stupid to read the fucking instructions. I imagine being able to type that out, and then have the AI remove profanity and not insult customers to be rather cathartic. Then, substitute "read the manual" for an actually complicated to explain thing.
"meant to survive rain storms" doesn't really mean the same as "meant to survive flooding", especially not "meant to survive flooding with salt water". For example, a car survives outdoor weather year-round, but you probably wouldn't want to buy one that was submerged in a flood zone. This plant seems to be floating, though, so they stay above water regardless.
Putting aside the salt water, you'll also just have lots of crud and debris on the panels if they get over topped, which then requires cleaning. Unless the rain in that area is strong and consistent enough.
Seems like a weird location to me, but what do I know.
I have experienced a positive return on investment from using Nix Darwin and devenv.sh since getting a new Mac two years ago. Did not spend too much time learning neither.
Perhaps I’m naive, but buying from an IKEA (in Norway) or another big store feels less risky than buying something handmade.
Several people are involved in making every product at IKEA. At least one of them must be an expert in compliance. They can expect scrutiny and product recalls, fines and bad sales if they’re found out.
The one person making the hand-made spoon does not necessarily know all the environmental regulations that should be followed.
I had a ceramic coffee mug that I loved and used very regularly. I bought it at a fair from a local potter and it had a very unique glaze, lots of blues and greens.
A friend who is a potter saw me drinking out of it and said that the glaze looked suspicious. He said it looked loaded with heavy metals, and that I should probably not use it or at least get it tested.
At the time I knew a guy who worked in a lab that tested certain substances for hazardous materials. He was intrigued and brought it in to work one day, and later texted me asking if I wanted it back, because it was very likely leaching cobalt, lead, and cadmium, and it was probably also very mildly radioactive.
I feel much better drinking out of mugs from IKEA and other big name stores.
Anecdotally, I find early mover advantage to be overrated (ask anyone who bought Betamax or HD-DVD players). It is significantly cheaper – on average – to exploit what you already know and learn from the mistakes of other, earlier movers.
reply