Hacker News new | past | comments | ask | show | jobs | submit
monster_truck | on: Muse Glimmer: 30B-parameter model optimized for always-on local agent workflows
Meta can never be redeemed, but it's still valid to admit that FB at one point had a very badass engineering culture.

They're one of 2 companies I would absolutely never work for (weapons etc aside). FB's recruiters hounded me so often I requested that they blackball me. The day they became Meta, I learned this by checking my email to see that they started trying to reach out again. I once again requested that they blackball me. This by extention taints OAI, the other company I'll never work for.

After a few hours with Glimmer I'm pretty impressed. It's better than the benchmark scores seem to indicate compared to Qwen 3.6 27B. I'm very excited for 3.8

swiftcoder | parent | next
> FB at one point had a very badass engineering culture

Perpetually kneecapped by one of the worst management cultures I've ever seen

fidotron | root | parent | next
I'm going there: Bosworth is the problem. He should never have got near being CTO.
MengerSponge | root | parent
Would you say those badass engineers were/are managed by Careless People?
swiftcoder | root | parent
My impression is that they had created a system where management cared very deeply about "number go up", and very little about "which number?"
MengerSponge | root | parent
The book title is a reference to The Great Gatsby: "They were careless people, Tom and Daisy- they smashed up things and creatures and then retreated back into their money or their vast carelessness or whatever it was that kept them together, and let other people clean up the mess they had made.” ― F. Scott Fitzgerald
aruggirello | parent | next
> It's better than the benchmark scores seem to indicate compared to Qwen 3.6 27B. I'm very excited for 3.8

Is it worth considering if it's only marginally better than Qwen 3.6 though? Qwen 3.8 27B is almost there, and will probably be better suited as drop-in replacement for 3.6. Not even considering there's probably going to be a 3.8-35B-A3B too - which will have even better performance.

petu | root | parent | next
Qwen3.6 is very token inefficient with it's thinking. Quantized versions often get into loops.

Glimmer is trained with 4 effort levels, not just thinking on/off. Maybe it's more token efficient in general. There's official 4 bit quantizations with reported 1% loss across 15 benchmarks -- so quants probably work good.

IMO that alone is worth trying for, even if they're otherwise equal.

monster_truck | root | parent
It takes 10 minutes to download and try, any model is worth at least that. In my experience benchmarks are generally dogshit
fidotron | parent | next
FB/Meta have had "don't recruit me" databases for a long time that are very easy to get yourself added to, thankfully.
LorenDB | parent | next
What is the other company that you would never work for?
zImPatrick | root | parent
He said OpenAI in the comment (if I read it correctly)
loading story #49243765
bko | parent | next
Meta doesn't need to be "redeemed". They have two of the most popular social media apps in the world. And theyll prob survive without ever having you work there
JKCalhoun | root | parent
I read that as, "Meta is a piece of shit but they're rich and don't care."

(Which I guess I agree with.)

loading story #49247196
gosub100 | parent
Mind telling me roughly what you had on your resume that had meta /fb hounding you for a job? ( Of course so I can avoid having this situation happen to me, naturally)
monster_truck | root | parent
Simply don't get flown out by IG to be their first mobile dev.