eurekin

Im a big AI/ML enthusiast (published one paper!) and was always flabbergasted to see scientists go off the typical provable/ testable lane and venture into philosophical and emotional territories

Oh, the static/dynamic split is brutal (and I believe some options have been moved around)...

Once you referenced routers, middleware and services simply by name, but that changed into per-source scoped versions (e.g. service1@file, middleware@docker).

I kept bumping into those edge cases (custom SSL cert set-up was really confusing), but thanks to chatgpt, I at least ended up with workable solutions.

Oh how incredibly convenient.

Plus, making ad blocking a channel owner's problem is kind of genius.

Bambulab A1 user here.

It's nice for simple stuff, but wouldn't recommend for anything needing precision. In the beginning it was laying out perfect layers, but after 6 months just lost it. Been doing every possible maintenance stuff and it just doesn't go back.

I'm personally betting, that they simply have easier access to more information, comparing to me at the same age.

When I started Uni, the "A diploma will guarantee you great job opportunities" mantra was unshakeable.

Now I think, the pendulum swung so hard in the other direction, that kids of same age have tons of refuttals at their disposal. It must take a lot more work, from parents, to instill and motivate what was once seen as a good career starter.

Certainly not defending LLMs here, don't mistake with that.

Humans do it too. I have given up on my country's non-local information sources, because I could recognize original sources that are being deliberately omitted. There's a satiric webpage that is basically a reddit scrape. Most of users don't notice and those who do, don't seem to care.

Kagi user here.

When I started using it (~ 2 years) , it was necessary. Google was simply not solving any of my actual issues (software related).

Now, It seems that google might have improved a bit. I check from time to time and the gap isn't as huge, as when Kagi started

We do. 0.5%

Plus that tens of terabytes of data you have to store for a week worth of traces

So how many generation T/s we can expect for a dense model?

I assume we can go up to 120B using fp8?

Not often article makes me want to buy shares. Like... never. This article is spot on.

Can concur. I jump between OnShape and FreeCAD in my free time. OnShape feels very polished. I go back to FreeCAD, because I've bought some models, I don't want to publish in OnShape's free tier.

It's amazing how much can be done, but anything that I think will take an hour, lasts 6. Many times into the night.

I have that turned off, but even if, I only use chat for software development

Oh, I went through that. o3 had the same memories and was always to the point.

My very brief interaction with GPT5 is that it's just weird.

"Sure, I'll help you stop flirting with OOMs"

"Thought for 27s Yep-..." (this comes out a lot)

"If you still graze OOM at load"

"how far you can push --max-model-len without more OOM drama"

- all this in a prolonged discussion about CUDA and various llm runners. I've added special user instructions to avoid flowery language, but it gets ignored.

EDIT: it also dragged conversation for hours. I ended up going with latest docs and finally, all issues with CUDA in a joint tabbyApi and exllamav2 project cleared up. It just couldn't find a solution and kept proposing, whatever people wrote in similar issues. It's reasoning capabilities are in my eyes greatly exaggarated.

I'm running a30-a3b-instruct q6 quant on exllamav2 and checked few simple tasks in roo and cline. Prompt adherence, tool calling and file changing worked flawlessly

Using tabbyApi and exLlamav2.

I'm still in awe that a local 3090 gpu was able to run the qwen3 coder instruct 30b-a3b exl3 q6 and...

Was able to create a sample page, tried starting a server, recognising a leftover server was running, killing it (and forced a prompt for my permission), retrying and finding out it's ip for me to open in the browser.

This isn't a demo anymore. That's actually very useful help for interns/juniors already.

Unusual for such outlets to take jabs at prominent companies. Normally, they are much more lenient. Interesting

I couldn't be more confused by this launch...

I had gpt-5 only on my account for the most of today, but now I'm back at previous choices (including my preferred o3).

Had gpt-5 been pulled? Or, was it only a preview?

Same here. I think it's one of those fields that feel polar opposite to, what they advertise to be.

Unfortunately, might not be easy in practice. It took years of effort to pass some certifications and audits and I don't think adding another language could be recouped now

I'm at a java shop. All customer facing apps are in java. Our tools and glue code is mostly in python. Similar for data processing workflows

Same here. Netflix, Hulu, Disney plus, HBO... I've had them disabled after months of zero usage.

YouTube on the other hand...

> chatbot instructed him to give up sleeping pills and an anti-anxiety medication, and to increase his intake of ketamine

That's 3 medications?

Also, how convenient those stories come out in light of upcoming "regulatory" safekeeping measures.

This whole article reads like 4-chan greentext or some teenage fanfiction.

I think this topic returns with some regularity... It often ends with justification about the need for a promotion of a particular executive that is involved with that inevitably undeniable success

> it returned 2021.1019.1.0, whatever that means

That looks like a version number...

Would like to see more of the captured data, because a simple "about" dialog, would also need to call some server to check, if it software is in the latest version. To display the "you have the latest version" label.

In my domain, our set of services only authorizes Customer Centre system to do so. I guess I'd need to ask them for details, but I always assumed they have checks in place

I received the e-mail exactly 15 hours ago:

   Hello,
   
   Your Pro plan just got way more powerful with three major upgrades previously available only to Max, Team, and Enterprise users. 
   
   Claude Code is now included
   
   Claude Code is a command line tool that gives you direct access to Claude in your >terminal, letting you delegate complex coding tasks while maintaining full control. You can now use Claude Code at no extra cost with your Pro subscription.

At work we dutifully delete all data on a GDPR request