I like this direction. The approach of letting LLMs figure out stuff on their own is flexible but inefficient and getting more structure can be very beneficial.

Well done for actually parsing the shell commands the LLM wants to execute. Another case where doing things “the old way” is both more correct and secure

This looks cool. Does it support working with local models?

Yes, you can use the dynamic provider to work with any local model you want (see scripts/providers)

It’s still capped at £1/kWh, which is a lot but also a price I’ve seen maybe for a few slots on a couple days in the last three years.

The price was known ~1 day in advance so I had the choice to fill up the battery with cheaper (but still high) prices overnight and reduce the impact

The framing of the article is very odd.

It says that Ubuntu increase the requirements not because of the OS itself but to have a better user experience when people have many browser tabs opened. Then it compares to Windows which has lower nominal requirements but higher requirements in practice to get a passable user experience.

I hope this means the years social media companies spent optimising for engagement are starting to backfire.

> A central issue driving the data is the changing nature of social media itself. The rise of apps such as TikTok and the popularity of video features including Instagram’s Reels mean that people are consuming social media more passively and are less likely to take an active role, a change compared with how they might have behaved on platforms such as Facebook.

Partition Magic was indeed magic. Thanks for the great work you did on it!

If you dislike the slop on the website so much you might also not like GitHub that much. It shows this software has been heavily co-written by Claude code

I'm aware of that! That's fine with me.

The model is only generating tokens without touching the network at all, right? How would it send data away?

Theoretically, by taking the opportunity to inject an exfiltration mechanism if you ask it to write code for you

Lots of people I know run models in "yolo" mode or the equivalent as well, which means it could just invoke curl or telnet to exfiltrate data.

GB News?

GB News gets a lot of criticism, but I watch the odd show and I've always found it to be balanced. There's a lot of political and regulatory pressure against them, so much so that GB News took Ofcom to the High Court and overturned illegal actions that Ofcom had taken.

At this point you’re doing almost as much work as if you handpicked a few channels and put them in your RSS reader.

Algorithms are sold as “curation is hard, the algorithm does it for you” but getting the algorithm to do a good job is actually a lot of work

No, I often get nice recommendations from the long tail, like interesting academic talks or tinkerer channels, videos sometimes with only a few hundred views or even fewer.