Your curated collection of saved posts and media
่ช่ก้ใๅ็ทใฎๆ ๅ ฑๅๆใซ๏ผก๏ผฉๆดป็จใธโฆๆตใฎๅ็ใ้ณๅฃฐใ่ชๅ่งฃๆใ็ตฑๅใ้ ็ฝฎใฎ็ซๆกใ : ่ชญๅฃฒๆฐ่ใชใณใฉใคใณ https://t.co/rMMNtdbJYQ ใ๏ผ่ช่ก้๏ผก๏ผฉๆดป็จใ๏ผๆตใฎๅ็ใ้ณๅฃฐใ่ชๅ่งฃๆ
่ช่ก้ใๅ็ทใฎๆ ๅ ฑๅๆใซ๏ผก๏ผฉๆดป็จใธโฆๆตใฎๅ็ใ้ณๅฃฐใ่ชๅ่งฃๆใ็ตฑๅใ้ ็ฝฎใฎ็ซๆกใ : ่ชญๅฃฒๆฐ่ใชใณใฉใคใณ https://t.co/rMMNtdbJYQ ใ๏ผ่ช่ก้๏ผก๏ผฉๆดป็จใ๏ผๆตใฎๅ็ใ้ณๅฃฐใ่ชๅ่งฃๆ
๐ Congrats to @nvidia on the release of Nemotron 3 Super โ day-0 support in vLLM v0.17.1! Verified on NVIDIA GPUs. 120B hybrid MoE, only 12B active at inference. Big upgrades over the previous Nemotron Super: - 5x higher throughput - 2x higher accuracy on Artificial Analysis Intelligence Index - Multi-Token Prediction (MTP) for faster long-form generation - Configurable thinking budget โ dial accuracy vs token cost per task - 1M token context window Supports BF16, FP8, and NVFP4. Fully open: weights, datasets, recipes. Blog: https://t.co/PAN0y778iB ๐ค Thanks @NVIDIAAIDev Nemotron team and vLLM community contributors!
Introducing NVIDIA Nemotron 3 Super ๐ Open 120B-parameter (12B active) hybrid Mamba-Transformer MoE model Native 1M-token context Built for compute-efficient, high-accuracy multi-agent applications Plus, fully open weights, datasets and recipes for easy customization and de
Interesting new paper on online RL for agents. Most agent training still treats deployment and learning as separate phases. Serve the model first, collect data later, fine-tune offline. But every agent interaction already contains a learning signal. This paper introduces OpenClaw-RL, a framework that trains agents from the next state that follows each action: user replies, tool outputs, terminal traces, GUI changes, and test results. The key idea is to recover two signals at once. Evaluative signals become scalar rewards through a PRM judge. Directive signals become token-level supervision through hindsight-guided on-policy distillation. In their personalization setup, the combined method improves the score from 0.17 to 0.81 after 16 update steps, outperforming binary RL or OPD alone. They also show gains across tool-call and GUI agents when combining process and outcome rewards. Why it matters? Agents should improve simply by being used. This is a practical step toward online, always-learning agent systems instead of static deployments. Paper: https://t.co/bzLdREoet8 Learn to build effective AI agents in our academy: https://t.co/LRnpZN7L4c
Sakana AIใฏใ้ฒ่ก่ฃ ๅๅบ ้ฒ่กใคใใใผใทใงใณ็งๅญฆๆ่ก็ ็ฉถๆใใใ่คๆฐAIๆ่กใฎ็ตใฟๅใใใซใใ่ฆณๆธฌใปๅ ฑๅใปๆ ๅ ฑ็ตฑๅใป่ณๆบ้ ๅ ้ซ้ๅใฎ็ ็ฉถใใๅ่จใใพใใใ https://t.co/3ZeIEqrK62 ๆฌ็ ็ฉถใงใฏใๅฝ็คพใฎๅผทใฟใงใใใๅฐ่ฆๆจก่ฆ่ฆ่จ่ชใขใใซ(SVLM)ใใ่ชๅพๅAIใจใผใธใงใณใๆ่กใๆดป็จใใใใญใผใณ ็ญใฎใจใใธใใใคในใใๅพใใใ่จๅคงใชใใผใฟใฎๅๆใป็ตฑๅใใใใฆๆ้ฉใชๆๆๆฑบๅฎใซ่ณใใใญใปในใไธๆฐ้่ฒซใง้ซ้ๅใใใทในใใ ใฎๆง็ฏใ็ฎๆใใพใใ ๅฎๅ จไฟ้้ ๅใซใใใใๆ ๅ ฑๅใใฎ้่ฆๆงใ้ซใพใไธญใๆฅๆฌ็บใฎAIไผๆฅญใจใใฆๆ่ก็่ชๅพๆงใ็ขบไฟใใๆๅ ็ซฏใฎ็ ็ฉถๆๆใๆฅๆฌใฎๅฎๅ จไฟ้ใฎๅบ็คๅผทๅใธใจๅฎ่ฃ ใใฆใพใใใพใใ
Sakana AIใฏใ้ฒ่ก่ฃ ๅๅบ ้ฒ่กใคใใใผใทใงใณ็งๅญฆๆ่ก็ ็ฉถๆใใใ่คๆฐAIๆ่กใฎ็ตใฟๅใใใซใใ่ฆณๆธฌใปๅ ฑๅใปๆ ๅ ฑ็ตฑๅใป่ณๆบ้ ๅ ้ซ้ๅใฎ็ ็ฉถใใๅ่จใใพใใใ https://t.co/3ZeIEqrK62 ๆฌ็ ็ฉถใงใฏใๅฝ็คพใฎๅผทใฟใงใใใๅฐ่ฆๆจก่ฆ่ฆ่จ่ชใขใใซ(SVLM)ใใ่ชๅพๅAIใจใผใธใงใณใๆ่กใๆดป็จใใใใญใผใณ ็ญใฎใจใใธใใใคในใใๅพใใใ่จๅคงใชใใผใฟใฎๅๆใป็ตฑๅใใใใฆๆ้ฉใชๆๆๆฑบๅฎใซ่ณใใใญใปในใไธๆฐ้่ฒซใง้ซ้ๅใใใทในใใ ใฎๆง็ฏใ็ฎๆใใพใใ ๅฎๅ จไฟ้้ ๅใซใใใใๆ ๅ ฑๅใใฎ้่ฆๆงใ้ซใพใไธญใๆฅๆฌ็บใฎAIไผๆฅญใจใใฆๆ่ก็่ชๅพๆงใ็ขบไฟใใๆๅ ็ซฏใฎ็ ็ฉถๆๆใๆฅๆฌใฎๅฎๅ จไฟ้ใฎๅบ็คๅผทๅใธใจๅฎ่ฃ ใใฆใพใใใพใใ
https://t.co/CHgEL0XrWP
we have now entered the timeline where i am using codex
https://t.co/CHgEL0XrWP
thereโs no timeline where apple releases something like this https://t.co/pGZRgXN7mV
me: "can you use whatever resources you like, and python, to generate a short 'youtube poop' video and render it using ffmpeg ? can you put more of a personal spin on it? it should express what it's like to be a LLM" claude opus 4.6: https://t.co/V9jCUk0CGZ
META has delayed the release of Avocado until at least May after it underperformed on internal evals, according to reporting by the NYT. They are considering licensing Gemini from Google as a temporary solution. https://t.co/mYfYN7uioN
excuse me? wtf @Microsoft ? https://t.co/vdBUuq8QdC
they made their "absolutely" theme claude coloured lmao https://t.co/HtZCorzeaB
Weโve been cooking. 2 updates in the Codex app ๐ You can now personalize the Codex app with themes that match your taste. Import themes you like or share your own. https://t.co/xOg9vzFxh1
they made their "absolutely" theme claude coloured lmao https://t.co/HtZCorzeaB
Time to make some mini groyps ๐ https://t.co/e7iMNUzQlg
You can now run Perplexity Computer directly from Slack. โ Install from the Slack App Marketplace โ Create workflows using context from your channels โ Everything syncs to Computer on web automatically No switching tabs. No copy-pasting. Just work where you already work. https://t.co/YT2h3VT9XB
Automations are now GA. You can now: โข Set the model and reasoning level โข Choose if runs happen in a worktree or existing branch โข Reuse workflows with templates Automations are great for recurring tasks โ daily repo briefings, issue triage, PR comment follow-up, and more. https://t.co/YlvQauuMqI
How many Codexes / Codices / Codii are too many? https://t.co/vH5GbHzWOf
How many Codexes / Codices / Codii are too many? https://t.co/vH5GbHzWOf
When this wins best editing on Sunday https://t.co/UiwIw7ilNe
When this wins best editing on Sunday https://t.co/UiwIw7ilNe
The Shield premiered today in 2002! This groundbreaking show was the first from our newly formed MiddKid Productions and starred Michael Chiklis as the corrupt, must-watch Detective Vic Mackey. Watch on Hulu: https://t.co/bzWTjWeAG9 @MichaelChiklis @CCHPounder @cathycryan https://t.co/AK8gEAQyOW
Anthropic's founders and employees are about to have a lot of cash. Many of them have pledged to give away huge amounts of that cash. But where's it gonna go? @cogcelia took a look in this excellent new piece: https://t.co/arIU9AjMAr
today is the only day you can rt this https://t.co/uhord5ZKcp
today is the only day you can rt this https://t.co/uhord5ZKcp
A guy just casually ate a lit cigarette in the background of a live FOX19 News broadcast in Cincinnati ๐ญ๐ https://t.co/Wf9hI04tCb
TinyFish + @n8n_io. Two things that just work, now wired together. TinyFish Web Agent is live as a community node. Your drag-and-drop workflows can now browse any site and return structured JSON. JSON workflow templates are open-sourced in the Cookbook. Import and go ๐ https://t.co/cVaCitq3C7
Groundsource is a new AI-powered methodology that transforms millions of public reports into a high-quality record of historical disaster data to aid crisis prediction โ starting with flash floods in urban areas. https://t.co/cjCvgi0h1m
New open source gaussian splatting repo, FriendlySplat launched yesterday. It compiles many nice features, ranging from pruning, TSDF generation, segmentation, and more! It comes with a Apache 2.0 license. Article: https://t.co/gsFNGJxWJd GitHub: https://t.co/qGZR6nMZqt https://t.co/OMuCVOHKQN
Morrissey when he wakes up and remembers he cancelled his show. https://t.co/fCHeQhSZac
Morrissey cancels show in Valencia โdue to sleep deprivationโ โIt will take me one year to recover. And that is an understatement" https://t.co/kE8LvaQMtu
Morrissey when he wakes up and remembers he cancelled his show. https://t.co/fCHeQhSZac
"became lifeless" new passive police language just dropped https://t.co/mkqglGfuax
Loved ones identify Innisfil man who died following police use of taser https://t.co/cimIugh6CM
