jvican

I use Adguard/wipr to great success. Why's uBlock Origin better? Also, wipr now has filtr, which provides URL filtering at the macOS level, so it's as efficient as it gets.

My pleasure. Make sure you check out the thriving ecosystem of tools around it as well. I’ve seen a few projects building on it in a prettt creative way.

I wanted to mention that `zmx` is a beautiful piece of minimal software that basically has gotten rid of my tmux and zellij use. Uses a ghostty-based terminal emulator for each session, supports pretty much all of these features by leveraging ghostty. I truly don’t see the benefit of something like herdr over this minimal, robust setup. If I really wanted a lot of customization for AI assistants, I’d consider using Orca, but my current zmx approach works OOB and it’s solid. Great software!

This is very interesting. I've only heard bad things about B70's performance with Vulkan, for example this Reddit link mentioning a Phoronix review: https://www.reddit.com/r/LocalLLaMA/comments/1sgdt7t/comment...

Similar to Input Mono font, which is superior to Monolisa in pretty much any way I can think of, plus free. https://input.djr.com/

Same here.

Nice! I didn't know about this company. Thanks for sharing. Has anyone tried and can they vouch for them?

Yes, and to be concrete, you can do so at economical prices at https://requestatest.com (it's a lifesaver in many occasions, I've used it 4 times with great success).

Much better than I originally expected. Drugs like kesimpta allow patients to self inject on a monthly basis relatively low doses, which limits side effects. Better than Ocrevus in that regard judging by what I’ve been hearing from fellow patients that have been using it for a while, even though equally effective.

There has indeed been an insane amount of progress over the past years. I didn’t know about stem cell therapy as an alternative to MS treatment.

They provide benchmarks as well. I like headroom’s approch a lot better. More transparent.

An ex colleague is working on Headroom, a much more legit alternative to RTK. They provide accuracy benchmarks in the repo and are transparent about the compression algorithms used for the different output types. I liked their approach a lot better than RTK and thought it might be relevant for you.

https://github.com/chopratejas/headroom

It's a good point. Though I do wonder if the magic he casted was more at the conceptual level (intense belief on a set of primitives that ought to work) more than the code itself. Even by 2018's standards, the Tensorflow code above doesn't really look that impressive. It's hard to judge based on those past standards, though. But, wonder if somebody who knows more than me can elaborate.

There's no doubt about Noam's abilities. But I read through that code, and struggle to see its 'magic' or 'alchemy'. Can you elaborate what you find especially good about that code? (You may assume GPU kernel programming knowledge on my end.)

The EU is large and most importantly very diverse. Pretty much all the West and South of Europe has a very strong small talk culture. You shall not stereotype a country, and even less so a political and economical union of countries.

I'm sorry, but this whole article just feels heavily AI-generated. The telltales are all over the place (the kinds of headline questions, the structure, the way math is presented, the super concise keyword salads, the bolding, the capital letters all over the place, etc.).

Do your own research when reading the information in this article, I wouldn't be surprised if there are hallucinations or incorrect information. FWIW, the author claims to work in this space.

Most of them can, but assignment 2 requires Nvidia GPUs as it heavily depends on you getting acquainted with Triton, Nsight compute and other GPU low-level programming tools.

I’ve always thought myself as a person suffering a bit of ADHD. I have ALL of your perceived symptoms. Making me reconsider how much on the spectrum I am.

Also, I think you should cut some slack to others people's comments about ADHD. You’d be surprised how many people don’t look like they have it but they actually do.

If you're interested in this resource, I highly recommend checking out Stanford's CS336 class. It covers all this curriculum in a lot more depth, introduces you into a lot of theoretical aspects (scaling laws, intuitions) and systems thinking (kernel optimization/profiling). For this, you have to do the assignments, of course... https://cs336.stanford.edu/

What a gem! Thank you for sharing.

Having an OS that versions all my VMs and allows A/B running the actual OS, has a high quality CLI, is built on modern standards, supports declarative files, and it’s simpler than Proxmox.

I don’t really care for enterprise support. Incus hits a sweet spot no other solution does.

I use my own solution (set of bash scripts) on top of IncusOS support for declarative files.

I use IncusOS in my homelab. It's a joy to set up and use.

Migrated from Proxmox and manage all my VMs. Heavily use coding assistants to automatically set things up through the IncusOS CLI, translate Docker-Compose images to Incus, write bash scripts to automate launching new containers to use `--dangerously-skip-permissions` without fear of repercussions, etc.

What I love the most about it is that it's possible to manage IncusOS with declarative files, so you always have visibility into networking setups, resource configuration, etc.

Highly recommend checking IncusOS out if you have similar use cases!

I personally like and use watchexec [1] to run commands on repeat.

Syntax example: `watchexec -r -e py uv run pytest -xvvs file.py`

[1] https://github.com/watchexec/watchexec

Plus, I don't think Japanese eat a lot more fermented foods than other cultures. It's way more prevalent in South Korea, China, Russia, etc.

I agree with you. I followed him out of curiosity for one or two months, watched about 10 of his videos.

He seems to have a good intuition, but he gives weak and often cherry-picked reasonings, to the point that many of his takes are completely unreliable.

For a channel called Predictive History, he made too many weirdly precise explanations and predictions that turned out to be wrong. Then, he'd look over the old failed ones to find new ones.

That being said, I'd say his macro level analysis is directionally correct, as well as his read on the incentives of each party involved. Watch his lectures, but be skeptical and double check everything he says, because he does indeed make factual mistakes... some of them are caught in the comments by other viewers, some are not obvious.

He might indeed need some personal development himself. I followed him when the US bombed Iran's nuclear sites last year. He was involved in a controversy with his kid and turned into a dick, going into a charade against the Western education system, for being overly harsh to his kid in a public space and getting reprimanded for it. I'm not condoning his behavior, I wasn't there, but I'd take anything this guy has to say on personal development with a pinch of salt. By the way, he published a post of apologies in his Substack IIRC.

Have you looked into zmx? [0]

It doesn't have built-in notifications and there's no panel to see all the open sessions, but I wonder how hard that would be to add.

I've used zmx since I ran into it a few weeks ago. Uses libghostty as well. It's great because it allows me to replace tmux completely in all my ssh sessions, and can keep one session per assistant.

[0] https://github.com/neurosnap/zmx

Does it do bufferbloat checks?

Is there any plan for this?

I used it for close to a year and abandoned it because I kept running into issues with tabs getting randomly reloaded and extensions causing trouble.

What would you say has changed over the past few months? I just felt like Kagi wasn't prioritizing Orion development enough, being busy with their main Kagi subscription and all.