jwr

Yes, they should function like gas stations and be regulated similarly. When I approach a gas station, I see the prices usually on large signs. If not, I always see the price I am actually paying displayed on a screen. None of these things are true for an EV charging station, which I think should be regulated.

Cheaper prices with subscriptions are not a problem at all. It's just that I should always see the final price that I will be paying per kWh before starting to charge.

On a related note, there is already a plug-and-charge system that identifies the car when it gets plugged in and is capable of charging the owner, but it almost never works. This is also something that should be mandated for larger EV adoption.

Thank you for Pandoc! I used it for many things over the years, and it was always there as a good option. I mostly use it for Markdown->Typst these days and it does this job very well.

No. And it's not the number of charging stations, it's the general shittiness of those. You never know how much you will pay per kWh, there are apps, networks, roaming agreements, cards, getting through this mess and starting to charge is always a challenge and takes a significant time (sometimes comparable to charging time!). You also get gouged on pricing, companies clearly set it so that the cost of 100km is comparable to that of a gas-powered vehicle. And for some reason people think EVs can only charge by backing into a stall, never in a drive-through bay like gasoline vehicles.

I own an EV and enjoy it a lot. I would not even consider buying a gas vehicle again. But while charging at home is great, charging during a trip is annoying. This needs regulation.

I hear so much about people running into token limits and using bajillions of tokens. Am I doing something wrong? I'm finding it hard to run into the limits of a Claude Max (20x) account. Well, perhaps the 50% Fable limit is something I have to be aware of, but otherwise this more than I can consume. I am the limiting factor: my capability to review code and design documents.

I use it to develop a large MRP-style app. This is as complex as things get.

Is it Clojure that makes things so efficient? My work style (lots of design documents)? I really don't know.

Do people skip the "review the code and read the documents" step perhaps?

What is the point of doing this? Why do people do this sort of thing? I'm genuinely puzzled.

I think the bigger problem here is not AI, but the increasing tendency for designers and designers-groups to look at one another's work and copy it. It almost seems like every designer has an inferiority complex where they look at others and think, "I'm not as good as them, therefore I have to mimic what they do."

This results in everyone copying Apple, for example, even their bad decisions.

Somebody noticed a while ago that all the AI company logos look like anuses — and they are right! They do!

Even product design decisions are not immune to this. Look at cars for example; Tesla started designing non-circular steering wheels and everybody followed. There is absolutely no reason to do this. A non-circular steering wheel in a road car is worse in every respect than a circle shape, yet everybody does it until the trend passes (hopefully soon).

If you look around, you will find plenty other examples.

Similarly here, I think that all AI apps have small icons not because they are AI apps, but because the design groups copy each other's work.

I test Qwen models regularly. They are very good for English and I'm guessing Chinese, but much worse for non-English (specifically, Polish).

In defense of this model, Gemma is actually a very good general-purpose model that can work with multiple languages. I use it for spam classification and for processing dictation, which means that I hold the entire model in memory all of the time, which is somewhat problematic (64GB RAM total, but heavy usage by docker, databases, etc)

I really think we should be solving a much bigger problem of the major email providers not providing an automated way of handling abuse and not caring about abuse reports at all. Most of my spam comes from the three major email providers and at this point I gave up even trying to send abuse reports because they just get ignored.

The big companies do not have to care because nobody will block Google, Microsoft or Amazon. They are too big to fail.

Spoofing a From field is an insignificant problem in comparison.

Incidentally, I've been testing a lot of dictation apps and models. I am on macOS Sequoia and intend to skip Tahoe altogether, so newer Apple models were not an option. I also dictate in multiple languages.

I mostly use the MacWhisper app until now, but I got frustrated by its slow development and lack of focus and dictation. So I tried the free TypeWhisper app. It works very well, but then I tested multiple models and my conclusion was that one cannot beat the Whisper Large v3 model for accuracy. I tried Qwen ASR (very good, but only for English), Voxtral (spoke Polish to it and it wrote Russian, BIG oopsie), Parakeet TDT 3 (very fast, but sometimes inaccurate).

I got frustrated by the speed of the Whisper Large model. So I spent an hour with AI benchmarking, testing and developing a new plugin. Turns out that on a modern Mac with a good GPU you can run the Whisper Large model at roughly the same speed as the tiny Parakeet model. 2-3x faster than what other apps ship. It's just that nobody bothered.

So, I am now happily dictating into TypeWhisper and then feeding the dictated text into a local gemma-4-26b-a4b-qat LLM for corrections. Everything happens instantly.

I find it unsettling that we are willingly outsourcing the decision on who can access our sites to an increasingly dominant corporate entity.

The reasoning behind this is also flawed: blocking "bots" and "AI" means that our AI agents working for us are unable to do their work for us, because of knee-jerk bot-blocks.

Those auto memories are also annoying because by default they are not version controlled and land in a secret location. I much prefer storing stuff in a version controlled `CLAUDE.md` file.

VW ID.4, and it's unlikely I can spare the time for making a video without a really good reason…

I have the same impression when I use things that are around me every day. Everything is broken and seemingly nobody cares. It seems that the teams developing software never use it themselves and do not care about bugs and terrible usability problems. It seems the management does not use any of the products that their businesses produce. And it seems that the ubiquitous breakage is just accepted as the status quo.

This extends to any device which has software or firmware: banking websites, banking apps on mobile phones, car firmware, etc.

I find it very depressing, especially when I think of the people who design, develop and deploy these apps. We are talking about large companies and teams of people. Entire chains of well-paid people designed, looked at, developed, tested, and then accepted these apps. And yet they break in the most obvious ways, or fail to be usable, instead pestering us with pop-ups and useless drivel and forgetting their state at every point.

These problems are not subtle. To take a glaring example from my everyday life, in my car there are buttons for moving windows up and down. Sometimes when you press the button to move the window down, it will actually start moving up. We are talking about a simple button that moves a window. And seemingly everybody thinks this is fine?

I don't know what to do about this.

This sounds a little bit like ADHD, but the common advice of getting medication isn't always applicable. Yes, it is definitely worth finding out if you have ADHD and diagnosing it, because if you do, it will help you understand yourself much better and might prevent problems further down the road like anxiety and depression. But medication isn't always the answer. ADHD might be entirely manageable for some people if their work is interesting enough.

Some things to try:

- noise cancelling headphones with white noise background (Iceland White or Naturespace apps provide really good quality natural recordings)

- clean desk

- clear split between work and non-work environments

Poland, while not entirely corruption-free, has managed very well after the transition from Russian-dominated communism. For reasons I still do not understand, organized crime was rooted out in the 90s, and gambling restricted to almost nothingness (gambling is always tied to organized crime).

Not perfect, but doing pretty well, and the effects can be seen.

Sigh. This kind of whataboutism derails all discussions online.

First, there is no requirement for "consistency". Every person chooses where to draw their moral line. The world is not black and white, but the unprovoked invasion of Ukraine by Russia, and subsequent (ongoing!) murdering, raping, and bombing of civilians is pretty clearly on one side of the line in my world.

Note that this in itself says nothing about the United States or any other country. You may consider every other country and choose not to support its economy, and this in no way makes the above point weaker.

That is a modern BBS! I love the idea. Though I guess most people these days don't know what a BBS is, if you don't, look it up.

Yes, please do. Your support of Russia is the only reason why I stopped using Kagi and paying for Kagi. I really want to like you and use your product.

Well it's a good thing they killed Aperture, then. Makes perfect sense.

(I was badly affected by that and decided not to trust them to maintain anything over the long term anymore)

Update: I now re-read the job posting and also realized that none of the responsibilities listed there are good for the users. This is bad news.

How is it better than, say, LM Studio? And does it run MTP models (which from what I understand are GGUF)?

The word "frontier" is like "load-bearing" at this point (Claude Code users will know what I mean). I wish we could stop using it. Especially as this does not, in fact, run the leading/top models locally on your Mac.

If you are developing something significant, you don't want your competitors to be able to later say "build me something like THAT over there".

I haven't measured the percentage of users, so I wouldn't venture an opinion with a number, but these concerns do exist.

Well, there is the small issue of privacy policy: Kimi will train their models on your interactions if you use their subscriptions, and only with direct API usage (billed at API prices) they say they won't. Whether you trust that is another matter.

Those things do make a difference to some of us, even though nothing is black and white. In my case, I'll probably want to wait until other providers appear through OpenRouter and then I'll try to judge how much I trust them. But even if I don't trust them much, they don't train models anyway, so the likelihood of my data being used that way is smaller.

As someone who has used both kinds over many, many years: it really doesn't matter.

I try to avoid "smart" appliances and if the manufacturer forces the "smart" stuff, I never enable it. Even my induction hob would like to connect to wifi (no idea what that would be useful for), but that's just not happening.

The problem with many cheap USB-C devices is that they lack a required resistor on the USB-C pins, which means they won't charge from a compliant USB-C PD charger, only through a USB-A to USB-C cable.

It's idiotic that companies don't even test this properly and never read the specs, but here we are. Many cheap Chinese devices like toothbrushes will have this problem.

Incidentally, companies developing technologies for reusing EV batteries in grid storage applications (where even <80% capacity EV batteries are just fine for many years), have trouble getting enough EV batteries, because they last much longer than we were made to believe.

Interesting. The major factor seems to be the latency of change-compile-test cycle.

I wonder if one of the reasons I've been getting such excellent results from LLMs with Clojure could be that they can immediately try things out in the running REPL. Rebuilding my entire application from scratch takes minutes, while connecting to a running system through nREPL and compiling single functions or namespaces takes less than a second.

I believe we are talking about the same thing.