M-x moix

471 posts

M-x moix banner
M-x moix

M-x moix

@chain_rules

ex-phd student (lasted 48hrs) | everything is a buffer™

Katılım Ocak 2024
253 Takip Edilen28 Takipçiler
M-x moix
M-x moix@chain_rules·
@ptr_to_joel no fucking way, I had to go an look if it was real
English
0
0
1
1.3K
Joel 🇦🇺
Joel 🇦🇺@ptr_to_joel·
holy wow they merged it
Joel 🇦🇺 tweet media
English
138
188
4.4K
818.4K
difficultyang
difficultyang@difficultyang·
bedrock is so ass it's unbelievable
English
9
2
82
7.3K
M-x moix
M-x moix@chain_rules·
@jackminong hold on, you might be onto something here
English
0
0
1
64
M-x moix
M-x moix@chain_rules·
@zodattack “it's not as bad as you think. it's not great either” we are so back
M-x moix tweet media
English
1
0
2
599
zöda
zöda@zodattack·
every software engineer i see is nervous about the job market. students are switching majors. twitter makes you feel like you should quit your job and start a company. I dug into the data to see the truth for myself. it's not as bad as you think. it's not great either. excited to share my findings & hear what y'all think:) link below.
zöda tweet media
English
16
5
157
29.5K
M-x moix retweetledi
vLLM
vLLM@vllm_project·
🎉 We just shipped a major redesign of recipes.vllm.ai. "How do I run model X on hardware Y for task Z?" now has a clickable answer. What's new: - URLs mirror HuggingFace: just swap huggingface.corecipes.vllm.ai in any model URL to jump straight to its recipe (e.g. recipes.vllm.ai/Qwen/Qwen3.6-3…) - Interactive command builder: pick hardware, variant, strategy (tensor, tensor+expert, or data+expert; single or multi-node; or a prefill/decode disaggregated cluster), toggle features → get the exact `vllm serve` command - Pluggable hardware: NVIDIA + AMD already integrated. One-click switch between Hopper/Blackwell and MI300X/MI355X, and the right flags and env are applied automatically - JSON API for agents: every recipe is also published at //.json (e.g. recipes.vllm.ai/Qwen/Qwen3.6-3…), so tools and agents can consume recipes without scraping - Contribute a new recipe end-to-end with the agent skill shipped in the repo: github.com/vllm-project/r… 🔗 recipes.vllm.ai Enjoy! ✨
vLLM tweet media
English
34
114
759
72.1K
M-x moix
M-x moix@chain_rules·
I think a lot of the ideas behind ipv4 tcp/ip can be applied to vLLM continuous batching lmao
English
0
0
0
7
M-x moix
M-x moix@chain_rules·
chud on-policy distilation aka real distillation aka hinton kd vs virgin off-policy distillation aka regular SFT
English
0
0
0
16
M-x moix
M-x moix@chain_rules·
In fact, after we've trained a model, I kinda do the opposite. I give it vague, dummy, often miss-leading prompts on my task
English
0
0
0
2
M-x moix
M-x moix@chain_rules·
I always avoid going down the prompt engineering rabbit hole, I've made the following rule of thumb: If you need to tweak your prompt for a model to work on your task it's because: a) Your model hasn't learned the task well b) Your task is not well defined
English
1
0
0
4
M-x moix
M-x moix@chain_rules·
@antirez HN Show has become obsolete due to AI, I used to love watching other people side-projects any idea on how you would approach this? is it even possible to have it back?
English
0
0
0
116
antirez
antirez@antirez·
HN shadowbanning is always cool. 6 upvotes in 17 minutes but no way it can reach the home page, while 4 votes in 25 minutes is there. Note that I don't ping any friend when I post, so all the votes I receive are spontaneous. Yet... Moderation system and broken algorithms are part of HN decline.
antirez tweet media
English
17
1
150
19.2K
M-x moix
M-x moix@chain_rules·
@thomasahle like figma mcp server being only available for those who pay the monthly fee?
English
1
0
0
33
Thomas Ahle
Thomas Ahle@thomasahle·
Jensen says tool makers, like Power Point, will be more valuable when AI agents 1000x their user base. But what prevents the agents from just copying those tools? We may see more tool makers create an agent moat around their tools. Your agents won't be able to use the tools directly, but have to talk to the PowerPoint agents. This is similar to many high paid professions today, like Lawyers or Doctors. They don't give you direct access to their tools, and that makes it harder to study and copy them.
Dwarkesh Patel@dwarkesh_sp

The Jensen Huang episode. 0:00:00 – Is Nvidia’s biggest moat its grip on scarce supply chains? 0:16:25 – Will TPUs break Nvidia’s hold on AI compute? 0:41:06 – Why doesn’t Nvidia become a hyperscaler? 0:57:36 – Should we be selling AI chips to China? 1:35:06 – Why doesn’t Nvidia make multiple different chip architectures? Look up Dwarkesh Podcast on YouTube, Apple Podcasts, Spotify, etc. Enjoy!

English
5
2
6
1.6K
M-x moix
M-x moix@chain_rules·
@SIGKITTEN “this is too dangerous to e” VC will contact u shortly
English
0
0
0
21
SIGKITTEN
SIGKITTEN@SIGKITTEN·
i cooked 2 crazy ass things today and i think they're too dangerous to even show
English
8
0
17
1.3K
M-x moix
M-x moix@chain_rules·
@ptr_to_joel suggesting andy pavlo “this year in dbs” blog post, not technical but funny read
English
0
0
1
19
Joel 🇦🇺
Joel 🇦🇺@ptr_to_joel·
the cmu student larp era
IS
1
0
4
281
Joel 🇦🇺
Joel 🇦🇺@ptr_to_joel·
will get lots of time next month so i might go database paper / oss diving and post about database stuff
English
2
0
8
470
M-x moix
M-x moix@chain_rules·
@SIGKITTEN Im so sorry, I hope the kids are taking this allright :(
English
0
0
1
18
SIGKITTEN
SIGKITTEN@SIGKITTEN·
goodbye, my sweet boy you were the first addition to our family 12 years ago my kids don't even know what life without you is yet our house feels empty we already don't hear your little squeak for food 3 hours before its time you wont come up to cuddle with the kids at bedtime anymore we wont find you laying in random boxes and suitcases anymore we wont hear your sister hissing at you for harassing her at 3am anymore we wont see the kids kissing you goodbye every day before school anymore rip Mouse
SIGKITTEN tweet mediaSIGKITTEN tweet mediaSIGKITTEN tweet mediaSIGKITTEN tweet media
English
54
2
226
5.8K
M-x moix
M-x moix@chain_rules·
does this mean I can use :oil from nvim onto S3 ?????
English
0
0
0
10