Hacker Newsnew | past | comments | ask | show | jobs | submit | ComputerGuru's commentslogin

How were you getting anything useful out of that? We found the (unquantized!) E2B model to be completely useless at even the simplest real-world classification tasks.

So what we want now is unsloth (or anyone) to release 4/6-bit quantized models of these releases?

Yep, Unsloth already did, as linked in the comment at the top of this thread

Why is this written as a marketing spiel instead of just, you know, answering the question?

ADBC: https://arrow.apache.org/docs/format/ADBC.html

Seems like a columnar version of ODBC, for OLAP instead of OLTP.


Quite aside from the architectural changes, I suppose this is the answer to why Google had such a glaring hole in the (pretrained) Gemma4 model lineup between the Gemma4 4b and Gemma4 26b models!

A model that comfortably fits in 16GB of VRAM (allowing room for context) is a welcome upgrade.


I’m not sure I agree? For each of your examples there are algorithmic approaches and neural network approaches. Companies have certainly been loose and wild with how they market these, but there remain distinct approaches and implementations for each. Very generally speaking, the neural network based approaches (aka “generative AI”) perform better but with much worse degenerative cases and a higher baseline rate of unwanted side effects (that are normally not immediately visible but tend to cause issues down the line).

My bigger concern is that these neural network based solutions have taken the place of the former rather than supplemented them. Many tools no longer provide the algorithmic/kernel-based approach at all, and have marketed the “AI” (née ML) alternative as a strict superset/upgrade, despite its potential drawbacks.

(Interestingly while the inference-based implementations generally have higher latency (or infinitely worse, cloud and pay-as-you-go requirements), for some computationally difficult kernels the inference-based approach is actually faster!


Microsoft has been releasing LLMs for years.

Sort of. Phi models were just trained on GPT outputs though.

For those that don't know about this. Phi was announced with a paper called "Textbooks are all you need". What they did was use GPT 3.5 and created synthetic textbook chapters and exercises.

They also did some more interesting work like showing very small models can be coherent as long as you have very simple children's book style training data (TinyStories is pretty famous).

Lots of these ideas are still used. Learning facts at scale with active reading is an ICLR 2026 paper from Meta AI that does a lot of similar work.


By design. The whole point of Phi is the "textbooks is all you need" theory on curated training data, as opposed to kitchen sinks.

They were mostly distilled or fine-tuned OAI models.

huh? The granite series isn't distilled

Granite is IBM

Ah snap. You’re right of course

And occasionally un-releasing them like with WizardLM.

Is this a Google-only proposal? Has Mozilla provided their thoughts on the matter?

When you have questions like this, the best places to check are the "standards positions" Github repositories for Mozilla and WebKit.

https://github.com/mozilla/standards-positions/issues/1369

Mozilla hasn't officially weighed in, but one decision maker (hsivonen) did vote in favor of it.

Apple WebKit is officially in favor of it, as of just a few days ago.

https://github.com/WebKit/standards-positions/issues/628


As far as I can see out-of-order streaming is only half the described functionality – there is also HTML streaming & revamped DOM parsing which does not have the positive signals that out-of-order streaming does:

https://github.com/mozilla/standards-positions/issues/1370

https://github.com/WebKit/standards-positions/issues/629


Yea we've been working with Mozilla and Apple on that one as well but they haven't responded to the standards position yet. I am sure that we'll reach something that's within consensus.

We've worked on the design closely together with Mozilla/Apple folks and with multiple external partners that want to use it.

Why would the default language-level union keyword implementation force boxing? Seems like a crazy decision if it could just also implement the HasValue/TryGet itself and avoid it?

Hi there! C# language designer here, and one of the people working on unions.

Boxing is not something inherently to be avoided. It actually can work better in many (most?) use cases, and avoids a lot of problems that non-boxing approaches often cause (like tearing and copy costs).

It's try that the non boxing pattern could be implemented by us. And it's very reasonable that that is something we may do post this release. However, it's a non-trivial area. There's no one correct 'non-boxed' implementation. For example, do you have separate fields for all your unmanaged data? or do you have a blob of bytes that is large enough to align all your unmanaged data from teh largest set of of unmanaged fields, and you unsafe index into that?

Similar question for managed data. Do you have strongly typed fields for that data? Or do you attempt to use objects, to compact to as little space as possible? The former avoids casting costs. The latter allows you to minimize space. You can also potentially use unsafe casts. But those might introduce memory holes in tearing situations. etc. etc.

Because of this, i think the best outcome is to define the pattern (which we've done) and then use generators to allow you to control precisely the impl strategy, giving you all the bells and knobs you want to best fit your domain.


Can we expect that the C# development team works with the F# crowd, to make this copied feature work on both?

Yes. We do. And unions should work well in F#. It's designed to be a very easy pattern for all CLR languages and compilers to understand (including F#).

It would help if generators usability and related docs would be improved, though.

They still feel like something only .NET team masters, and a few MVPs that wrote key reference blog posts about them.


This one keeps crashing the browser after it reaches 100%. Safari/iOS, iPhone 13. I was able to navigate and use a few of the other ones linked to from the comments, though. Curious.


I have the same phone and it crashed for me too after reaching 100%


Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: