cptcobalt

This is really exciting—the 12-32b size range has my favorite model size on my home computer, and the mistrals have been historically great and embraced for various fine-tuning.

At 24b, I think this has a good chance of fitting on my more memory constrained work computer.

It's also just harder to support because of the sheer fragmentation.

I give my engineers different Android phones as their primary development devices and yet the weird Android issues that keep cropping up are near constant.

As an adult with too much disposable income and a 3090, it just becomes a local LLM server w/ agents when I'm not playing games on it. Didn't even see the potential for it back then, but now I'm convinced that the xx90 series offers me value outside of just gaming uses.

Anecdatally (& disregarding political SMS), my spam call/text frequency has substantially skyrocketed to about 2-4x that of around 2020, to the extent that I am tempted to change phone numbers.

This is a nifty idea. I bet there'll be a HN-correlated pop in fridge whiteboard sales from this.

Below 50v is considered "low voltage". Higher voltage would require different safety considerations.

It would be even cooler if Amazon also encouraged and built a "Kiosk mode" browser view of the kindle for this sort of display hacking.

You can easily use a plugin like https://www.continue.dev/ and configure it to use Claude 3.5 Sonnet.

This article reads with the tone & pacing of AI slop.

Edit: coverage with better reporting & screenshots of the problem: https://www.theverge.com/2024/9/30/24258693/ps5-update-news-..., https://www.ign.com/articles/ps5-homescreen-now-replaces-uni...

This article reads with the tone & pacing of AI slop.

I think this counts as flight time, not paperwork & preparation time on the ground.

This is the best description of orchestra tuning I've ever read. I've always wondered what gives it that specific sound of everything coming into place with relative consistency across orchestras, but didn't really know the mechanics of it. Thank you.

> Also, is it viable to produce something like the A320 in low quantities when airlines need fleets in the hundreds?

Bombardier thought so, and did so! They developed and introduced the C-series aircraft to compete with smaller 737s and A320/A319. After introduction, Boeing fucked them over so hard that they sold the aircraft program after introduction to Airbus for a token sum. Airbus now builds and sells the C-series as the A220.

https://en.wikipedia.org/wiki/CSeries_dumping_petition_by_Bo...

I'm in Tier 4, and not far off from Tier 5. The docs aren't quite transparent enough to show that if I buy credits if I'll be bumped up to Tier 5, or if I actually have to use enough credits to get into Tier 5.

Edit, w/ real time follow up:

Prior to buying the credits, I saw O1-preview in the Tier 5 model list as a Tier 4 user. I bought credits to bump to Tier 5—not much, I'd have gotten there before the end of the year. The OpenAI website now shows I'm in Tier 5, but O1-preview is not in the Tier 5 model list for me anymore. So sneaky of them!

The puberty of web engines!

I'm soon to be building an updated gaming PC (while ship of theseus'ing parts like my 3090, SSDs, etc), and it looks like this Zen 5 series is...not it. My sights are likely set on the 7900X3D as the processor I'm picking.

Interesting how much this undersells the product. I really thought I wanted it when I saw Avi's tweets about using his prototypes, where (for instance) it was coaching him on delivery and tone during a meeting. This proposition—just as a virtual friend—feels extremely on the nose in this era of tech, like we're just inviting "Her" to happen.

I'm still keeping my preorder, but I'm a bit more cautious.

They're unrelated, but the typeface is intended to evoke the vibe of computers in that era.

> your car’s movements and exact location are being collected and shared anonymously with a data broker

> said that the company had gotten the precise location of about 10 million G.M. cars up to every one to three seconds, but that the data did not include identifying details about the driver.

The word "Anonymously" means exactly nothing here. If one's exact location and movements are collected, wouldn't that include a lot of time and datapoints where the car is spending time at work and home locations?

Seems incredibly trivial at that point to deanonymize usage.

I wanted more examples of trolls and how the behavior degrades. The one scripted joke example doesn't land with a punch.

> In short, I believe that the revealed objectives of LLM vendors differ starkly from what they say. They pour considerable resources into responding to every trollish LLM transcript that goes viral on the internet — and it’s useful to ponder why.

I'd be pretty interested in reading at how people do poke at LLMs and what this actually reveals about the vendors priorities.

This is HN, is asking about a VPS a bit quaint, when we're in the era of container orchestration/etc?

Really, hidden motors in the tour? That would be an interesting read.

I think this is true as we keep building up abstraction layers. Computers are getting faster yet feel slower as we just want to work with higher level tech which makes it easier to understand less of how the sausage gets made.

But I don't think this is a now problem, in the age of AI, but has been a growing problem for decades as software has matured.

Anything that is a response to classic "SEO manipulation" to get higher ranking on a search engine result page to create an appearance that the content is higher value, is more comprehensive, or took more effort to produce does create net-slop. And that's been going on for 10+ years.

This document is delightfully (?) 90s-era MS Office coded, with plenty of clip art and word art. Somewhat funny to see.

For a rather dense document from 2004, is any of this worth casual study and still relevant?

Their users would revolt, it's a point of pride for B-unit holders to show off how well they can use seemingly inscrutable UIs.

This unironically feels like good research & paper potential.

This is over trivializing it, but there isn't much more inherent complexity in training an 8B or larger model other than more money, more compute, more data, more time. Overall, the principles are similar.

If I saw more packaging like this in my local stores, even if slightly more expensive, I'd pick it up every time. I think the overall friction is more in the return process, and that's where the space for innovation really lies.