alex000kim

7 ms·

1.
Show HN: NanoRL – RL training for LLMs in ~1,800 lines(github.com)
11 points by alex000kim 2mo ago|0 comments
2.
Scheduling jobs across Slurm clusters (and K8s, and cloud) from one place(skypilot.ai)
2 points by alex000kim 2mo ago|0 comments
3.
RL Is Bottlenecked by Inference. Scale It Independently(skypilot.ai)
11 points by alex000kim 2mo ago|2 comments
4.
by alex000kim 2mo ago
No, Uzbeks themselves use “non” when using Latin alphabet. Fun fact: in the past Uzbek language has been written in arabic, cyrillic, and, since ~2018, in latin. Source: grew up there
5.
by alex000kim 3mo ago
same experience last year, existing without having to stop was especially surprising
6.
Silverback Imfura took a chance, and ended up alone(gorillafund.org)
75 points by alex000kim 5mo ago|24 comments
7.
Do AI Detectors Work Well Enough to Trust?(chicagobooth.edu)
5 points by alex000kim 5mo ago|1 comments
8.
You've been doing harness engineering all along(alex000kim.com)
7 points by alex000kim 5mo ago|0 comments
9.
Lessons from Going Solo(alex000kim.com)
7 points by alex000kim 6mo ago|1 comments
10.
by alex000kim 6mo ago
That directory is huge already! I guess the index.md helps the agent find what it needs, but even the markdown file is very long - this would consume a ton of tokens. Also I wonder who/what decides what papers go in there. In the blog
11.
by alex000kim 6mo ago
yup, as the blog says > The full setup works with any project that has a benchmark and test suite. so having a clear and measurable verification step is key. Meaning you can't simply give an AI agent a vague goal e.g. "improve
12.
by alex000kim 6mo ago
I am sure this would works well in general. There is a challenge wrt to how to make them communicate effectively to e.g. 1) avoid duplicative work and 2) allow them to combine/overlay each others' findings to yield even better res
13.
by alex000kim 6mo ago
sounds similar to "LLM Knowledge Bases" https://xcancel.com/karpathy/status/2039805659525644595
14.
by alex000kim 6mo ago
technically you're correct, but look at the prompt https://github.com/alex000kim/claude-code/blob/main/src/util... it's written to _actively_ avoid any signs of AI generated code when &quo
15.
by alex000kim 6mo ago
Oh right, I just saw https://news.ycombinator.com/item?id=47582220 will update the post with this link
16.
The Claude Code Source Leak: fake tools, frustration regexes, undercover mode(alex000kim.com)
1376 points by alex000kim 6mo ago|578 comments
17.
by alex000kim 8mo ago
Author here. I've seen the docs you linked to: Slurm uses "gang scheduling" to mean something specific (timesliced oversubscription where jobs alternate on shared resources). I'm using the term in its broader CS sense: a
18.
by alex000kim 9mo ago
This was so clearly LLM-generated that I couldn't get through the whole thing.
19.
by alex000kim 1y ago
I created this PR to make it easier for folks to train and serve it on any cloud (or their own K8s): https://github.com/karpathy/nanochat/pull/18
20.
The Evolution of AI Job Orchestration. Part 1: Running AI Jobs on GPU Neoclouds(blog.skypilot.co)
2 points by alex000kim 1y ago|0 comments
21.
Bulk Object Storage data migration with SkyPilot(nebius.com)
5 points by alex000kim 1y ago|1 comments
22.
Orchestrating LLM Fine-Tuning on Kubernetes with SkyPilot and MLflow(alex000kim.com)
2 points by alex000kim 2y ago|0 comments
23.
ML experiments in the cloud with Skypilot and DVC(alex000kim.com)
3 points by alex000kim 3y ago|0 comments
24.
Why Sales Engineers Exist(alex000kim.com)
2 points by alex000kim 3y ago|0 comments
25.
Don’t know what to do next? Teach(alex000kim.com)
2 points by alex000kim 3y ago|0 comments
26.
Show HN: GPT4-powered Slack bot that can scrape URL contents(github.com)
8 points by alex000kim 4y ago|3 comments
27.
How to Create an Outstanding Data Science Portfolio(medium.com)
3 points by alex000kim 5y ago|0 comments
28.
Building your Data Science career(youtube.com)
1 points by alex000kim 6y ago|0 comments
29.
by alex000kim 7y ago
Kind of. Fastai abstracts a lot more of Pytorch then what keras does to TF.