RTX Spark does go up to 128gb No idea how it compares tho

I have one. The limitation (beyond total size of the memory) with the Spark is DDR5. "Real" inference hardware is HBM (high bandwidth memory) which is like 10x the performance.

So for prefill -- which is more about compute than bandwidth -- the Spark performs quite admirably. But on decode it's highly bandwidth constrained. Some smaller MoE models (e.g. Gemma4) can do 60-70 tok/second but anything dense, or anything that is actually filling up most of that 128GB is going to choke out around 15 tok/sec. Even at NVFP4.

My experiments with this are at https://github.com/rdaum/eider/

For my current work I get to log into trays on a real GB300. It's somewhat comical that NVIDIA is marketing the little baby on my shelf here as even in the same universe as that. Which is basically like having access to a super computer.

Yep, it's mostly garbage, I feel like I woke up with hangover honestly.

No AI writing in the book so far (but that does not mean it won't be garbage anyway lol)

Well I take back what I said then and best of luck to that specific endeavor.

Pixel Art Flight Sim https://www.youtube.com/shorts/6jderLN3Aqk

I don’t know about the Linux kernel, but I’ll be surprised if don’t have some “fully vibe coded OS” for Christmas (which would be cool to see)

I recall seeing a claim about a vibe coded os already on Reddit somewhere. Looked very windows 3.1 but didn’t investigate further

For db just give it credentials of a readonly user, for instructions you can do this. You can give setup a list of approved tools and bash commands https://www.anthropic.com/engineering/claude-code-best-pract...

Do you let it consume PII? Anything related to authenticaion?

Not everyone is handling PII. Where I work, anything like that is only available to a very limited set of people who absolutely need to be able to see it. Also some systems allow access control at the column and even row level, so even if it's intermingled with other data you want the LLM to read, you might be able to mask it that way.

Also, people shouldn't be running any LLM on data of a business without a proper contract in place like you have with any vendor who has access to your data. And if there's specific PII requirements, those should be covered too.

A system-agnostic language for magic spells with a compiler capable of producing the magic wand movements, incantation, hand signs or magic circle required to perform the spell

Is this so you can automate the spell casting via robotics?

I know it sounds a lot like 2025, but it's actually to create a dataset of spells for training my wizarding AI :P

(Original comment was not a joke btw, it's part of another thing I will publish eventually)

Py = 3.14 :')

/compact does most of that, for me at least

/compact we will now work on x, discard y, keep z

the trouble with compact is that no one really knows how it works and what it does. hence, for me at least, there is just no way I would ever allow my context to get there. you should seriously reconsider ever using compact (I mean this literally) - the quality of CC at that point is order of magnitute significantly worse that you are doing yourself significant disservice

When you hit ^O after compact runs (or anytime), it tells you exactly what compact did, so it isn’t that mysterious

if you actually hit the compact (you should never be there no matter what but for the sake of argument) more often than not you'll see CC going off the rails immediately after compacting is done. it even doesn't know what it did let alone you :)

You mean you never stay in a CC session long enough to even see the auto compaction warning?

100% - read this - https://blog.nilenso.com/blog/2025/09/15/ai-unit-of-work/

WAN Show is going to be wild this week

My most productive day (so far)... How AI is changing workflows of my pet projects