I'd say most subscribers to the channel want to watch his snarky ramblings as they are entertaining. I'd say the information provided is secondary to the format and personality.
I for instance don't subscribe because I want to know about old toasters, christmas lights and the internal combustion engine. If that was the case, I'd subscribe to channels specific to that topic. I subscribe to Technology Connections because I like his videos, and if I learn a thing or two along the way, that's a nice extra
I use the Street Complete app, which shows missing data around me. It's fun and satisfying to add many details that no other map has, like accessibility of crosswalks or street lighting.
Sorry for the tangent, but how does Qwen3.8-flash-next compare to DeepSeek v4 Flash? I still haven't found the time to set it up, but I'm really happy with DeepSeek v4 Flash
What offends me the most is how they used visitor's compute and badnwidth for ther own purpose unrelated to the reason the visitor accessed the site, without consent or disclosure. It's MY computer, MY internet, not yours to do with it what you please
I set something similar up in a project and it's been surprisingly good. In my case I ignored all top-level except for specific files and dirs such as Dockerfile, src or tests. Juniors rarely (never?) need to add new files or dirs at top-level, and seniors tend to (always?) realise they need to explicitly stage that new dir/file. Other prejects that have the usual exhaustive .gitignore generating from mixing and matching multiple templates ended up with random .claude or .zed once people started using new tools that weren't around when the initial gitignore was created
IMO, if you depend on AI to make a living, I'd invest in hardware for local inference, and learn on how to effectively make a living using AI inference you control, on hardware you control. Sure, economically speaking it's way cheaper to use one of these heavily subsidised services (for now), and their models are faster and more capable, but if your livelihood depends on AI inference, and you are renting AI inference, you are a being a serf of the tokenlord. And your livelihood depends on the whims of the tokenlord. They can increase rent prices, they can decide you can no longer do whatever you are doing, and you have no recourse, because you are dependant on them to make a living.
In their page they say it will need at least 112GB[0], so including context, that would be a tight fit. I'm also hoping I can make a q4 fit on my 128GB strix halo
I'm saying this based on anecdotal observations from my biased POV, so take this with a grain of salt. There are income differences, but most are inter-city/region differences. I'd say within a specific city or region, it's low gini, but there's a high gini if you compare cross-region. Income and costs in a village in rural Spain are less than half that of a "mid-tier" city, and less than a third or even quarter of that of Barcelona or Madrid
Try out Snikket. Really easy to set up, with sensible defaults. You just docker compose up and have a pretty funcional XMPP server without needing to configure much. Adding complex things such as gateways onto Snikket is a bit more complicated, as you have to deal with Docker volumes etc to put the right file in the right place, but I found it easy to manage. I've been running my Snikket instance with gateways for WhatsApp and Matrix for a couple of years now and I'm really happy. It's very low maintenance
the lesson I take from this whole saga reinforces my convictions and what I was already working working towards: tie data to physical locations, where we still have rights. Self-host everything at home. Portable devices should always have the bare minimum data needed for the time until back home, and not even a way to connect to the data at home from a portable device outside home. You want my data? get a warrant for searching my home, and it's there waiting for you
This! The problem is not whether AI makes more or less mistakes than a human. But for decades, people have been used to computers either working, or crashing, but never working wrong or misleading. AI changes that, and people really need to understand that. But that goes against the interest of AI provider's and their investor's interests, so the point is not being transmitted to the end users prominently enough.
And what's the right way to use Gemma? Where can I find the correct template and settings if those aren't the ones provided by Google, Unsloth, and aren't built into llama.cpp? I discarded using Gemma 4 because it got into weird loops when tool calling
> My only issue is getting it in hands of some crazy dictator, but that is another story.
That IS the story though. The kind of offences these cameras disuade pale in comparison with the risk of abuse of power infringed in the 99.999% of innocent people these cameras are pointed at suppose.
There are two companies that have lost my trust, probably forever: X and Meta. I don't see myself ever trusting anything coming out of either of these companies ever again
Agreed. I bought my low profile keychron k3 pro, I believe it was, several years ago and I'm really happy with it. No complaint whatsoever. I bought the equivalent with full layout for my wife a couple of years ago and she's really happy with it too.
I don't think I can move away from low profile mechanical keyboards with programmable firmware like QMK now. I'm too used to my Keychron
He (and I) might have been wrong on how useful the LLMs could get, but time isn't proving him wrong on his claims that the usefulness doesn't justify the investments around LLMs
The worst I've seen changed the view out a window from an alley to a private garden. Most also add natural light that isn't possible, and in some cases I'm convinced the generate image is depicting the space as larger than it actually is, with furniture and spaces that wouldn't fit in reality
I might be wearing my tinfoil hat too tight, but I think that market pressure will continue prioritising shipping worse code faster, especially now that with AI it's possible to ship code really fast. The "solution" to expensive hardware will be way less capable hardware and we'll end up with "thin clients" and all compute will happen in centralised "clouds". I hope I'm wrong
do you find Qwen3.5-122B to be SOTA-level? I moved from it to Qwen3.6-27B (both Q8), and I prefer 3.6-27B, and it leaves me room to spare for other small models
I've been dailying GrapheneOS for more than 2 years. I haven't had any issues with any of my bank apps. Although there are apparently some that do block usage on non-google-certified OSs. The only limitation in my use cases pre-GrapheneOS that I've found is not being able to make NFC payments through Google Wallet, but I found a bank in my country whose app implemented NFC payments, and made an account with them to be able to convenintly pay with my phone.
edit: and I'd like to add, GrapheneOS brought me back the joy of using my phone. Since 2018 or so I started to dread my phone (and the internet) more and more. Installing GrapheneOS brought back the joy on using these marvelous computers (and self-hosting brought back the joy of using the internet)
what I keep wondering is what kind of laws will be rendered useless with the precedent they'll cause. Can this be beginning of the end of copyright and intellectual property?
if it were true that indeed was legal to rewrite and relicense open source code, would that also be true for non-open source code? as in, could someone do a similar rewrite of their employers proprietary code and release it publicly?