einarfd

I’ve been building sandboxing for Claude code workloads. So I can let it run wild without breaking my computer. Originally I used docker, but I’m now in the process of jettisoning that, and switching to qemu.

For my use case I want ssh access and being able to use docker in docker. This allows for things like test containers and docker compose. You can get all of that working with docker. But you kind of have to fight docker the whole way.

NanoClaw might have different needs, and docker could work better for it, and I hope so for their sake. But I’m not optimistic.

Could it be that USA government believe that Iran might be trying to do something similar to the Ukraine operation spiderweb, where they attacked the Russian long distance bomber fleet with short distance FPV drones? While there aren't bombers at Fort Bliss. As far as I know there are other high value targets.

I see a lot of people here are worried that we will end up with ads in all AI vendors products, or at least the frontier labs. I think this is unlikely.

We are already seeing a market for AI for productivity in companies, the Claude code product is the first serious one here, but we can expect more to show up. When you look at the B2B market, ads are basically not a thing in these segments, companies are generally more willing to pay for products, and less willing to accept outside influence on how the product works, and I don't think this will change when companies are buying AI either. Companies might be happy with selling ads in their own products. On the other hand consumers, don't like to pay, and that will probably drive consumer oriented products to be ad funded. Basically what I'm expecting will happen, is that we will end up with two types of AI vendors.

Those that target the consumer market and those that target the business market. Consumer AI will trend toward companionship, entertainment, casual chat — things like digital friends, relationship play, even adult content. Companies want none of that, and some of it is serious legal liability. Even a few missteps and you get expensive backlash in the business market.

It does look like OpenAI is trying to succeed in both the consumer and business market, and there are companies that are able to pull this off, most do not, and end up serving one of the markets. Given their lead in the name recognition I suspect they are going to end up an ad financed consumer brand, and will lose the business market to someone else. But I might be wrong.

The saving grace for those of us that don't want ads to bleed into our AI tools, is that we probably will be able to buy the same products that the small business segment buys. Some consumer oriented features might be missing, but they might either be features we don't need, or maybe open source could fill the gaps?

Colors and color names are culture dependent, and you are not guaranteed that people in different cultures agree on what color something is.

The most famous of these discrepancies is Japan and green vs blue, or why does Jenkins by default use red, yellow and blue instead of red, yellow and green.

So I would urge using something other than colors as an example of shared human experience.

My iPhone calls it personal hotspot.

Not only isn’t this or any other of the DMA features accessible in the USA, but Norway which is a member of the EEC and which therefore both have to and is in the process of ratifying the DMA. Don’t get this either.

That Apple is so petty that it blocks on legal technicalities like that, when everyone knows it is just a matter of time. Really sours me on the whole company.

When I used to be on call for Cisco WebEx services. I got paid extra, and got extra time of. Even if nothing happened. In addition we where enough people on the rotation, so I didn't have to do it that often.

I believe the rules varied based on jurisdiction, and I think some had worse deals, and some even better. But I was happy with our setup in Norway.

Tbh I do not think we would have had, what we had if it wasn't for the local laws and regulations. Sometimes worker friendly laws can be nice.

The quotes I've been seeing from him, makes him sound like a cartoon villain, I'm hoping for his sake he is misquoted.

The current Russian regime believes in sphere of influence, and they view Ukraine, the baltics, at least some parts eastern europe, and probably more as a natural part of their sphere of influence.

This crashes with the western view where countries and populations have a right to self determination. Some of the countries that Russia want to fall under their sphere are also members of the EU, which make this even more problematic. Seen from Moscow, EU and western countries have encroached on their turf and this is a problem for them. Seen from the western side, this is wrong, and if Russia is such a bad neighbor that its neighbor join defensive alliances to get out from under their thumb, that is their own fault, and the way the world is supposed to work.

Russia also has a geographic vulnerability where there is no geographic chokepoints from at least Poland and straight to Moscow, which historically has given Russia problems.

Give this, there is actually a rational for what Russia is doing, personally I think it is a bad rational, but there is logic in the madness, even if from my perspective, it is based on a deeply wrong world view.

A bit off topic, but I find the tone of voice of ChatGPT with GPT5 really off-putting. While the antrophic models or even Le Chat. Sounds much better to me. While the examples used in this article aren't that bad, OpenAI has to be more that just marginal better, to make me want to use it, when it "talks" like that.

In Norway the plates to owner database is public, and you can either use the free government service or for profit ones. To look them up.

Basically ownership of vehicles is not viewed as private information. So it isn’t covered by privacy legislation. But tracking where the car is, that is covered. That doesn’t stop, speed traps reading car plates, nor parking tracking with cameras. But that info can’t be sold for other purposes.

This is a great idea, and worth doing. An other option in Claude code, that can be worth trying, is the planning mode, which you start with ctrl+tab. Have it plan out what it's going to do, and keep iterating it, until the plan seems sound. Tbh. I wish I've found the planning mode earlier, it's been such a great help.

What I use my digital id for, is services, where the provider needs to know that I am me. I only use a small part of the services that we have in Norway that you can access with out digital id (BankID) solution, but those are useful for me, and I do not think all of them would exists without it.

For governmental services, I use it for things like logging into health care services. Where I've used it for checking my prescriptions, and communicating with my doctor. If I had kids I would have used it for contact with the school. An other governmental use is tax filling and tax returns which comes around every year, and this is just scratching the surface.

When it comes to non governmental usage, it is mostly bank and bank adjacent usage. I do use it to log into my different banks, my stock broker, and insurance providers.

The solution we have in Norway, is not perfect and one of the persistent problems, are that not everyone can get one, and since it is used a lot by the government, not having it, makes you a bit of a second class citizen. I do believe that they are finally doing something about that, and that the system will be redone a bit next year, so even if the banks don't like you. You will be able to get one.

In my opinion the Norwegian variant of this, is an important part on making it easier to switch things like banking or insurance provider. Without BankID I think the competition between these type of providers would be even worse.

The privacy story of this looks better than the Norwegian BankID approach.

I would like if Norway moves in this direction, and I think that through the ongoing alignment with the EU wide program on digital id, that might happen.

If you look at the countries that are lauded at having the best online government services. They all have some type of digital ID.

Having something like that is imo. a cornerstone for building out top notch digital governmental services, and I don't fault the UK for trying to get this in place.

That being said, I'm not convinced it will be that much of a blocker for illegal workers. I'm sure they will find a way around it.

Oops, seems the quote is an old one, and not news. That invalidates my original post somewhat, and I'm sorry that I didn't do proper due diligence.

Here is the original post:

That doesn't sound like the rhetoric of someone who is winning. It sounds more like something someone pushed into a corner, and seeing their project crumbling would say.

But bringing up that it is about civil liberties is an important point, not the way he would like though.

You would think that trying to keep the discourse about criminals and pedophiles would be smarter for his side? I do not follow Danish politics, but I do start to wonder if he is just not very good at doing politics?

If you look at the article, they do not use his name. Even after there is a final judgment against him. He only got 60 days of probationary prison, while I personally found it weird he was the only one that got punished. I'm sure the punishment could be a lot worse.

On the name side, the names of the people involved where sparingly if at all mentioned in the press. The did use the names of ministers and top military and navy officers of course. But not the crew. I'm sure their peers knew who they where, and that some careers got hurt. But if you left the navy and did something else, it probably wouldn't follow you though.

Driverless taxis is IMO the wrong tech to compare to. It’s a high consequence, low acceptance of error, real time task. Where it’s really hard to undo errors.

There is a big category of tasks that isn’t that. But that are economically significant. Those are a lot better fit for AI.

This seems fine to me.

Having these models terminating chats where the user persist in trying to get sexual content with minors, or help with information on doing large scale violence. Won't be a problem for me, and it's also something I'm fine with no one getting help with.

Some might be worried, that they will refuse less problematic request, and that might happen. But so far my personal experience is that I hardly ever get refusals. Maybe that's justs me being boring, but that does make me not worried for refusals.

The model welfare I'm more sceptical to. I don't think we are the point when the "distress" the model show, is something to take seriously. But on the other hand, I could be wrong, and allowing the model to stop the chat, after saying no a few times. What's the problem with that? If nothing else it saves some wasted compute.

When reading the article I was reminded of reading Sarah Wynn-Williams book Careless People. The carelessness and disregard for obvious and real ramifications, of the policy choices of management, seems to not have changed from her time at Facebook.

If they didn't see this type of problem coming from mile away, they just didn't bother to look. Which tbh. seems fairly on brand for Meta.

That they exempt politicians is basically admitting that the security problems that detractors bring up is true, and is something that should be used against them.

After all exempting some police, that work on investigating child molesting, from the scanning, that is understandable.

Exempting prime minster Mette Frederiksen, on the other hand. Means either that they understand that it undermines security, or that she or some other top politicians are child molester. So which is it?

Depends on what business OpenAI wants to be in. If they want to be in the business of selling AI to companies. Then "firing" the consumer customers that want someone to talk to, and double down models that are useful for work. Can be a wise choice.

I have gpt-5 on my iPhone, but not on my iPad. Both runs the newest chatgpt app.

Maybe they do device based rollout? But imo. that's a weird thing to do.

I got the same when trying it with standard gpt5. But when I used the thinking mode I got:

3 — cranberry.

Tried with Claude sonnet 4 as well:

There are 3 r’s in the word “cranberry”:

c-*r*-a-n-b-e-*rr*-y

The r’s appear in positions 2, 7, and 8.

I would expect standard gpt5 to get it right tbh.

There are companies out there working on those problems as well. How the funding climate for them are. I don't know. But the market for smart robots, should be gigantic. So there must be some. Keep in mind that what is easy, and hard for a human, which is the result of billions of years of evolution. Isn't necessary the same things that are hard or easy for our technologies.

And deleting the file would have been perfectly fine, if Tesla had given out the file when asked for it. Instead of what they did.

You are right, but unfortunately you are the least useful right, which is technical right.

That is definitely what auto pilot means in the aeronautical and maritime sphere.

But a lot of the general public has a murky understanding of how an auto pilot on a ship or a plane works. So for a lot, probably the majority of them. They will look at the meaning of those two words and land on that auto pilot, means automatic pilot. Which basically ends up beeing self driving.

Sure in a perfect world, they would look up what the term means in the sphere they do not know, and use it correctly, but that is not the world we live in. We do not get the general public, we want, but we have to live with the one we got.

Are you sure that there aren't multiple tiers of rental skies in resorts in the USA? I have friends in the rental business in Norway, and they rent out skies from the kind of beginner friendly skies that would fit your description. To top of the line Stockli and Van Deer kit. I think they keep all of their gear in at least decent shape, with the top gear kept in the best shape, as it makes the most difference for those customers. Btw. haven't rented much in the Alps. But my impression is that you can rent gear for all ability levels there aswell, and that that gear is kept in good shape.

Power is why you would build in Iceland. There is a lot of geothermic potential there. It's after all, the land of hot springs and volcanoes. So if you need more power, build a new geothermic power plant.