jwr

> One reason to strongly prefer Apple over Google is that (as I understand it) you can still walk into a store with ID and recover your online identity.

Is this true? If so, in what countries? (Poland has no Apple stores, for example, for inexplicable reasons Apple has been giving the middle finger to this market)

That would be a huge difference.

How would moving your number to a new phone help? We're talking TOTP here, not SMS "auth".

> Small fender benders and sideswipes are much more frequent

Do you have any data on this? Everything I've ever read about them points to exactly the opposite conclusion. They actually reduce accidents and by a significant margin.

My only guess would be that you are referring to the US, where people are not used to roundabouts and yielding without a STOP sign.

Sorry, it was a thinko — I meant "solar panels".

Inverters are indeed expensive, so are batteries (and you will need batteries to make good use of solar; otherwise, you will be getting the ridiculously low energy company rates for what you over-produce). Being far north is somewhat compensated by the fact that solar batteries run much more efficiently when they are cold. Also, modern solar cells are surprisingly effective on cloudy days, unless it is really dark.

Complaining on HN :-)

More seriously, I have not found a solution.

Every ad system I tried resulted in nearly 100% bot traffic to my site. Google and LinkedIn give you "tools" that will show you various numbers, but I implemented my own conversion tracking, because I didn't trust their numbers.

I decided to stop burning money on these ad systems. I'm not sure if they work for anyone at this point, I'm guessing this is more of everybody being convinced they might work and trying to get them to work, rather than ads actually resulting in conversions.

Well, I guess this might be different if you are an online casino or something similar.

I find it rather sad that the original posts have been posted on an open site, where anyone can read them, while the HN post points to an increasingly dubious site that doesn't even show the post unless you create an account and log in.

I thought the spirit of the open web would be more important to this community.

I always thought it was enough to switch off the "Improve the model for everyone" setting on chatgpt.com:

"Allow your content to be used to train our models, which makes ChatGPT better for you and everyone who uses it. We take steps to protect your privacy."

But apparently there is also an entire completely different route "Do not train on my data"?

Does this mean that before I submitted the "Do not train on my data" request, my data was used for training in spite of "Improve the model for everyone" being turned off?

We are getting to facebook/meta-levels of privacy settings obfuscation.

I also have a lifetime pass, but I'm becoming increasingly annoyed with Plex. Why do I have to regularly turn off their "recommended" stuff that is just ads for movies I can buy online? Why does it appear and re-appear?

Also, the Plex app for LG TV has been buggy and broken for years: some content will not play at all, some will not play if you ff/rewind, and some will not play until you physically turn your TV off and then on. Very annoying.

I'm glad to see competition.

I have an LG TV. It has been pestering me with those licensing agreements from time to time, but interestingly enough, I was only required to agree to fairly reasonable basics, everything else I just ignored and left for later.

I'm guessing this is an oversight and the TV does record, but then this is criminal. But, nobody will bother to prosecute them I guess...

I gave up on reporting spam. Almost all of the spam I get (and I get a lot of it!) comes from Google, Amazon and Microsoft. None of them care much, none of them make reporting easy, and it seems they don't follow up, because the flood of spam never ceases.

They are "too big to fail": few people can afford to just block Google. This is why I am very happy to see people starting to do just that.

Acceptance rate is good, but MTP doesn't help in my case because of my machine's memory bandwidth constraints (M4 Max). Turn out it's better to turn MTP off.

I would (tentatively) agree — but that doesn't invalidate my argument: qualia are not objectively measurable, so you still can't tell if a machine is conscious.

If there is any company that can "ban your account" which would result in big problems for you (because, for example, they host your E-mail which means you've just lost access to password recovery for every account you have) with no recourse or due process, you should seriously rethink your approach to your online life.

Host your own E-mail, people, or use reliable E-mail providers that are not huge corpo-behemoths that sell you to advertisers. Those smaller E-mail providers will have actual support that you can contact in case of problems and will not "ban you".

I think before we get to self-referentiality, we have a bigger problem to solve, and that's one that I did not expect: namely, at which point do we stop saying that a machine pretends to be conscious and start saying that it is conscious?

We have, somewhat unexpectedly, built machines that are very, very good at pretending. Now, I'm not calling the current generation of LLMs we have "conscious," but I can't really define a marker or a boundary beyond which I would start calling them "conscious."

For what it's worth, I've spent some time with Claude to develop a local runner for `llama.cpp`. I run Qwen3.6-35B-A3B-MTP (fast!) and Qwen3.8-27B (20 tokens/s).

This was definitely worth the effort. Benchmarking and testing various approaches and various options really paid off. For example, one thing that surprised me was that MTP made things slower, not faster for Qwen3.8-27B.

I use a 64GB MacBook Pro (M4 Max).

I got so tired of every hardware device requiring cloud services, accounts, logins, privacy policies, etc. that I started building my own. I think the tipping point was a phone app to control a Panasonic air conditioner. It could just communicate with the air conditioner over Bluetooth, after all, the air conditioner was never more than 3 meters away from me. But instead, Panasonic chose to build a cloud services platform, so you need internet access, you need to create an account with an appropriate password, agree to data processing, accept the privacy policy, accept privacy policy updates, etc.

Once you do all that, your spouse comes and asks for a way to control the air conditioning too…

This whole "you need to create an account" is a virus that needs to go away. It should be possible to control a local device without accounts, privacy policies, and other crap.

I spoke English and Polish to it. It produced Russian or Ukrainian as output.

From what I understand, they also have no certifications (CE/FCC/VCCI and similar) and cannot really be sold as consumer products, only as kits.

I used MacWhisper for a long time and then recently switched to the free TypeWhisper. Both work very well on a Mac.

It's interesting that the results can be so different depending on the person, the use case and even the microphone used. I use dictation a lot, so I try to stay up to date with the latest models as much as I can. So far, for my needs, nothing could beat Whisper Large v3. I keep hearing that Nvidia Parakeet models are better, but they just don't work as well for me, even though they are unquestionably faster.

Things change a lot if you need to speak to the model in multiple languages. There are very few models out there that can automatically detect the language spoken and produce correct output.

I tried the larger Voxtral models, but they didn't work for me at all. When I spoke to them in Polish, they produced output in Russian or Ukrainian.

For now, I settled on creating my own plugin for TypeWhisper, which runs Whisper Large on the GPU and does it much faster than pretty much anything else out there. But I'm still hoping that something better will come along, as Whisper Large is quite old at this point.

I use cLaude cOde in my eMacs (through ghostel, which is unbelievably great!) to generate code :-)

I then use mAgit to review it.

The problem is when you don't want to send your data to any provider. Private documents, health information, all that is something you don't want to send anywhere, even if the provider makes a pinky promise that they won't use your data for training or store it.

And no, I do not use gmail, whatsapp or instagram :-)

Shipping with their computers -- maybe not, but they did make FoundationDB open source (again) and are developing and maintaining it.

Hmm. What would be really helpful for me would be an enhancement of the TypeWhisper app for dictation that would be able to get the context of what I am dictating into and send it along with the prompt to an LLM.

Probably much simpler and much less of a privacy problem (I run my own local LLM for that purpose so that nothing leaves the machine).

Unfortunately, paying money to an economy of a country which started and still wages an unprovoked war against another country, killing civilians indiscriminately, is a choice. The only reason you can "focus on building great software" is because you are not in Ukraine, where Russian bombs paid for by the aforementioned Russian economy would be falling on your head.

Not to mention the problematic nature of trusting any data coming from a country known for state-sponsored deception, trolling and organized operations in informational warfare.

(to the inevitable whatabouters: yes, this can also be applied to other economies/countries, everyone makes their own moral choices on where to draw the line, but the line should be drawn somewhere)

I am saddened by your stubborn stance on this.

Ghostel is fantastic for this. It is very fast, fully compatible, and you can quickly enter a copy-paste mode with C-c C-t. With ace-jump this makes for very efficient copying.

Since I started using Ghostel, my Claude Code and Codex sessions started gradually migrating into Emacs. They work just as well as in a terminal and managing them is more convenient. This is also why I can't really find any use for all those newfangled tools that are intended to help with managing multiple coding sessions.

I have been using Emacs for 32 years now. It had its ups and downs: there was a period of very slow development, which led to the XEmacs fork, which I used for many years.

These days I see a renaissance of sorts. First, the development progress is excellent. Second, Emacs turned out to be the ultimate material for AI. I no longer write my own Emacs functions. I tell AI to modify the Emacs behavior the way I want. I now have a development environment that I fully control. Windows open where I want them. Keystrokes do what I want them to do. I have tools for annotating Claude Code plans in Emacs and plenty of other functionality. Without AI, I would never be able to implement all these tiny fixes and improvements. It would cost me too much time. And in other IDEs, I would have to "ask for permission", because not every behavior is exposed and modifiable. In Emacs, almost everything is, and you can change the behavior of other packages, too.

It feels great to have a development environment that I am in control of and that won't disappear tomorrow, like so many other IDEs did.

This is… not obvious. It's only easy if you want the simplistic approach of sorting lexicographically or numerically. But in more complex tables, you will often encounter cases where the data representation for sorting is different from data representation for display.

Of course most JavaScript-based tables get this wrong, too.