HackerTrans
TopNewTrendsCommentsPastAskShowJobs

mezark

no profile record

Submissions

[untitled]

1 points·by mezark·il y a 13 jours·0 comments

What happens when you run a CUDA kernel?

fergusfinn.com
294 points·by mezark·il y a 15 jours·32 comments

A running list of reasons to move to open source

whyopensource.ai
6 points·by mezark·il y a 22 jours·0 comments

Moe inference optimizations: 15% lower expert load by request reordering

blog.doubleword.ai
3 points·by mezark·il y a 2 mois·0 comments

[untitled]

1 points·by mezark·il y a 2 mois·0 comments

Tensor Network Attention

mainlymatmul.com
2 points·by mezark·il y a 2 mois·0 comments

Redundant Information in LLM Weights

fergusfinn.com
5 points·by mezark·il y a 2 mois·0 comments

Tans: Precomputing RANS

fergusfinn.com
3 points·by mezark·il y a 2 mois·0 comments

Also-RANS: Asymmetric Numeral Systems for Entropy Coding

fergusfinn.com
25 points·by mezark·il y a 2 mois·0 comments

70x faster cold(ish) starts for SGLang

fergusfinn.com
4 points·by mezark·il y a 3 mois·0 comments

QueueSpec – drafting speculation tokens while a request queues

blog.doubleword.ai
1 points·by mezark·il y a 6 mois·0 comments

ZeroDP: Just-in-Time Weight Offloading over NVLink for Data Parallelism

mainlymatmul.com
1 points·by mezark·il y a 6 mois·0 comments

Parallel Primitives for Multi-Agent Workflows

fergusfinn.com
1 points·by mezark·il y a 6 mois·0 comments

New fastest AI Model Gateway – 450x less overhead than LiteLLM

github.com
2 points·by mezark·il y a 9 mois·0 comments

Should GPUs Make Free Trade Agreements?

doubleword.ai
3 points·by mezark·il y a 10 mois·1 comments

comments

mezark
·il y a 12 jours·discuss
really like the framing of this post :)
mezark
·le mois dernier·discuss
I love this blog
mezark
·le mois dernier·discuss
(As someone who cares a lot about philosophy of consciousness / & cogsci)

The whole point of consciousness being a 'hard problem' is that we just cannot make claims like 'X is not conscious'
mezark
·le mois dernier·discuss
we think so - but haven't tested it ourselves
mezark
·le mois dernier·discuss
Hi! Co-founder of Doubleword here - we've hugely increased the number of models that we offer (partly thanks to work that we've done on hotswapping https://blog.doubleword.ai/fast-sglang-starts.

We're kind of known for our low prices - our prices (our main usage is for our high throughput API - the async tier) is significantly below average openrouter prices - but cached prices is coming soon which will lower them even more :)
mezark
·le mois dernier·discuss
We at doubleword are bullish for AMD for low-interactivity inference - it does just take a bigger lift on the software side...
mezark
·il y a 2 mois·discuss
If you're talking about UK sovereign LLM inference you need to mention Doubleword... very serious inference optimization lab in london with public endpoints for OS models
mezark
·il y a 10 mois·discuss
We look at how comparative advantage from economics applies to LLM inference - some GPUs are relatively better at FLOPs, others at memory bandwidth. What happens if you let each do what it’s best at?