Yampeleg Profile Banner
Yam Peleg Profile
Yam Peleg

@Yampeleg

Followers
39K
Following
28K
Media
3K
Statuses
15K

The only AI researcher they sent a missile for ๐Ÿ‡ฎ๐Ÿ‡ฑ | Co-host @thursdai_pod โ€ข AI news every Thursday

Joined July 2012
Don't wanna be here? Send us removal request.
@Yampeleg
Yam Peleg
4 months
Lore TL;DR: > spent summer in SF > missile hit my place back home > officially refugee, returned, officially displaced > snuck into collapse hazard building, rescued GPUs > patched together a working setup > technical posts coming back to x .com > apologies for the inconvenience
16
9
344
@Yampeleg
Yam Peleg
1 day
miss the simple fp32 life
1
0
6
@AckeeBlockchain
Ackee Blockchain Security
11 days
We just shipped the first VS Code extension built specifically for @Solana developers. Static analysis detectors that catch vulnerabilities in the IDE + fuzzing coverage visualization. The missing layer between development and audits, integrated into devs' workflow โ†“
9
22
89
@Yampeleg
Yam Peleg
2 days
PewDiePie went from Minecraft let's plays to building 10xGPU rigs, training 120B LLMs, distributed vLLM inference, coding webapps, RAGs, speech recognition pipelines, linux sysadmin.. Millions of kids are watching him compile CUDA now. 110M+ subs. WHAT AN ABSOLUTE LEGEND!!
173
1K
21K
@Yampeleg
Yam Peleg
2 days
Just found ๐š๐šŽ๐š–๐š’๐š—๐š’-๐šŒ๐š•๐š’ is goated for literature reviews Ask an open research question, give it a little push toward arXiv, and it goes off batch calling its ๐—š๐—ผ๐—ผ๐—ด๐—น๐—ฒ๐—ฆ๐—ฒ๐—ฎ๐—ฟ๐—ฐ๐—ต tool, reading papers, multi-step reasoning with 1M context.. It finds crazy insights
2
5
86
@Yampeleg
Yam Peleg
2 days
bro woke up on halloween and chose trick
0
0
3
@Yampeleg
Yam Peleg
2 days
god forgive me for all the lies i tell llms
1
0
13
@Yampeleg
Yam Peleg
3 days
plz stop pretending you run โ€œ1000 agents in parallelโ€ so the rest of us can code thx
0
0
28
@Yampeleg
Yam Peleg
3 days
this is insane even for huggingface wow
0
0
2
@Yampeleg
Yam Peleg
3 days
this is insane
1
0
2
@Yampeleg
Yam Peleg
3 days
wow this whole thing is even interactive and also explain details about all the different LLM architectures, this is above and beyond!!
1
0
3
@Yampeleg
Yam Peleg
3 days
hf are doing godโ€™s work fr
@_lewtun
Lewis Tunstall
3 days
We've just published the Smol Training Playbook: a distillation of hard earned knowledge to share exactly what it takes to train SOTA LLMs โšก๏ธ Featuring our protagonist SmolLM3, we cover: ๐Ÿงญ Strategy on whether to train your own LLM and burn all your VC money ๐Ÿชจ Pretraining,
4
6
126
@Yampeleg
Yam Peleg
3 days
LIVE: ๐Ÿ“† ThursdAI - Halloween Special! | Weekly AI News: SCARY FAST Agents (MiniMax M2, Cursor 2.0, SWE 1.5), OpenAI's ASI Plan, 1X NEO Humanoid
1
1
8
@Yampeleg
Yam Peleg
4 days
Iโ€™ve been saying this for years: Instead of causal attention, do: N ร— N ร— N = O(Nยณ) โ–ฒ โ–ฒ โ–ฒ โ”‚ โ”‚ โ”” fully connected attention โ”‚ for every token up to current token less efficient training but insanely more powerful with same inference cost as today O(Nยฒ)
@snowclipsed
snow
5 days
just out of vain curiousity ; what happens if you increase the complexity of attention? like, has anyone tried cubic attention lol
26
19
437
@Yampeleg
Yam Peleg
4 days
every time i see an emoji in a github readme i just assume itโ€™s a grift and close the tab
2
0
8
@Yampeleg
Yam Peleg
4 days
man i can actually count on codex itโ€™s wild told it โ€œmake my training code fasterโ€ left it in yolo mode controlling 8x A100s and went to lunch lmao
2
0
9
@Yampeleg
Yam Peleg
6 days
wait why not use the full logits actually?
4
0
8
@Yampeleg
Yam Peleg
6 days
TL;DR: โ€ข SFT: Mimics teacher but learns only from teachers states โ€ข RL: Explores on its own but gets sparse reward Solution: 1. Student generates trajectory 2. Teacher scores each token (logprobs) 3. Reward = -KL(ฯ€_ฮธ || ฯ€_teacher) per token 4. Policy gradient update Wow nice
@thinkymachines
Thinking Machines
6 days
Our latest post explores on-policy distillation, a training approach that unites the error-correcting relevance of RL with the reward density of SFT. When training it for math reasoning and as an internal chat assistant, we find that on-policy distillation can outperform other
20
32
514
@Yampeleg
Yam Peleg
10 days
Claude Code just politely asked me to stop coding
1
0
25
@Yampeleg
Yam Peleg
10 days
LIVE: ๐Ÿ“† ThursdAI | Weekly AI News | ChatGPT Atlast Browser, DeepSeek OCR, Claude Code Web, Browserbase Director
0
0
8
@Yampeleg
Yam Peleg
12 days
OpenAI shipping social networks, browsers, app stores, video gen, code editors, agents, frontier LLMs, open weights models, CLI tools... ...is NOT because some AI is vibe-coding like crazy behind the scenes. It's because their people are good.
@danshipper
Dan Shipper ๐Ÿ“ง
12 days
In less than a month @OpenAI launched a totally new social media app / format AND a new browser. Can anyone remember a time when companies this size were shipping so fast? Itโ€™s a new benchmark for the velocity of world class software teams.
1
2
26