Ned

Ned

@nedinthecloud.com

Tech educator, runner, podcaster, music fan, ๐ŸŒฎ devourer HashiCorp Ambassador and Microsoft MVP Find my things: https://nedinthecloud.com

945 followers662 following594 posts

Long one today. Ramping up for a marathon in November. Found a friend in the middle of the path ๐Ÿธ and I shooed them to the side.

The 666 on the CD is from my days of working at Hot Topic. We had to label our CDs so they wouldn't get mixed up with everyone else's collections. I think we were only supposed to play CDs supplied by the company. We, um, did not do that.

Ned reposted

I feel terrible asking this, but my debit card got stolen and I donโ€™t currently have the funds in my account to cover some upcoming payments while I sort the dispute out with the bank. If anyone could help spot me, I would really appreciate that! V: @Ian-Coldwater CA: $iancoldwater PP: @coldwater

Client: we love what you did! Also client: now can you just make it shorter, change the graphics, and redo everything? Me: yes, yes I can ๐Ÿ˜ญ

Sometimes a sponge falls in a sink. Sometimes it lands on the drain. Sometimes your washing machine uses that sink. And sometimes your laundry room floods. ๐Ÿ˜ž

Overall a disappointing and lackluster #VMA. Snoop was phoning it in, everything skewed heavily Gen-X, and there were barely any top tier artists there. MTV seems to be accelerating their slide into irrelevance.

Ducking wut? Of course, it is the VMAs after all. Congratulations to Taylor for making the most out of the least. #VMA

Sorry sombr you're no George Michael. Thank goodness Raye and Adam Lambert were there. #VMA

So um, does MTV know something about Madonna that we don't? They're giving her all the flowers ๐ŸŒธ #VMA

I guess since MTV knew Taylor wasn't gonna win any regular awards this year, they might as well make one up.

My wife and I often refer to fitting everything in the refrigerator as fridge jenga. Technically it should be fridge Tetris, but I always mess it up. Weird.

I have an event on my calendar for October 1st called "Did the AI bubble pop?" I don't remember when or why I added this.

Basically everything about running open weight models is a slog. Projects are out of date, versions are incompatible, and the docs are... well they aren't great. Feels like everything is moving too fast to be stable and we're all tao dancing on quicksand.

Clearly I've been listening to too much DCC. Every time someone writes in all caps, I can't help but hear it in the voice of Princess Donut. ๐Ÿ‘‘๐Ÿ™€

I do recommend getting the flu and COVID shot ๐Ÿ’‰. I do not recommend planning to do anything else the next day.

Today's adventure was in using llama.cpp to save a trained model as a GGUF file and then quantize it. Turns out using transformers v5 to train led to compatibility issues with the scripts in the llama.cpp repo. Gotta love the cutting edge.

I know intellectually that Nvidia GPUs have better software support and a more mature ecosystem. But watching a five year old 3090 beat out an AMD card from last year by a sizeable margin? There's knowing and experiencing. Just wild.