SUN YOUNG HWANG α―
π°π·
@SOSOHAJALAB
Followers
727
Following
23K
Media
368
Statuses
7K
CEO, AI (LLM/LMM/MLX) Engineer and enthusiast with non-tech and construction field based. Two kidsπ§ββοΈ, Fencerπ€Ί, F1 loverπ but love tech more π»β€οΈ
Seoul, Republic of Korea
Joined May 2023
Our domain with AI solutions!. - Vision solutions for defective of finished materials object detection with gpt-4-vision and object detection tasks. - Answer the questions for defective of finished materials from GPT, gemini and our own model locally. - OpenAI GPTs for our own
3
0
13
Let me try this!! π.
Introducing Devstral Small and Medium 2507! This latest update offers improved performance and cost efficiency, perfectly suited for coding agents and software engineering tasks.
0
0
0
Another weapon on this?.I might cancel others now π
Weβre rolling out more updates to Claude Code. First, the GitHub Actions integration is now available to Pro and Max plan users (in addition to pay-as-you-go API users).
0
0
0
Another super paper for memory LLM π.
Brilliant Memory framework proposed in this paper. MemOS makes remembering a firstβclass system call. LLMs forget stuff fast and retraining them costs a fortune. MemOS treats memories like files in an operating system, letting a model write, move, and retire knowledge on the
0
2
9
I might subscribe the grok 4 heavy!. This must be better than claude π
0
0
0
Possible to ship π°π· π€.
Opening orders for Reachy Mini today, our open-source desktop robot for AI builders, starting at $299! Fully integrated with @LeRobotHF & @huggingface for the whole community to build AI apps for it (like this dancing one). We'll probably ship a first batch of a hundred this
0
0
1
RT @AndrewYNg: New Course: Post-training of LLMs. Learn to post-train and customize an LLM in this short course, taught by @BanghuaZ, Assisβ¦.
0
330
0
Finally!! Grok 4 is coming real soon π.
0
0
1
The Godsloth @UnslothAI.
We made step-by-step guides to Fine-tune & Run every single LLM! π¦₯. What you'll learn:.β’ Technical analysis + Bug fixes explained for each model.β’ Best practices & optimal settings.β’ How to fine-tune with our notebooks.β’ Directory of model variants. π
0
0
4
I need this π€.
Hugging Face just dropped Reachy Mini. an expressive, open-source robot designed for human-robot interaction, creative coding, and AI experimentation. Fully programmable in Python (and soon JavaScript, Scratch) and priced from $299. Reachy Mini measures 11β/28cm in height and
0
0
0
Huggingface has done it always π€π€.
We just released the best 3B model, 100% open-source, open dataset, architecture details, exact data mixtures and full training recipe including pre-training, mid-training, post-training, and synthetic data generation for everyone to train their own. Let's go open-source AI!.
0
0
1
Interesting paper for context consistency π.
MemAgent. MemAgent-14B is trained on 32K-length documents with an 8K context window. Achieves >76% accuracy even at 3.5M tokens!. That consistency is crazy!. Here are my notes:
0
0
1
Another solid method for LLM memory π.
0
0
0
Vibe coding is still going up π.
Wild. Emergent 2.0 ships full stack apps from a prompt. Now with Google Auth, LLM integration, auto security & scalability reviews. Vibe coding will never be the same. 10 wild examples + how to try:. 1. Veo 3 Ad Wizard
0
0
0
Midjourney is cool as veo3 π€.
My mom sent me a 30-year-old drawing. That I made for her when I was a kid. Naturally, I animated with Midjourney. It's not perfect. But it's awesome. And it captures the innocent chaos . Of my childhood imagination. Yes, my mother is fighting a dragon. With a
0
0
1
Unstoppable AI models on lmarena π.
3 new models live in the Arena today π . π§ Mistral Small 2506: latest 24B open model (Apache-2.0), tuned for efficiency by @MistralAI.π¨ Imagen 4 Ultra: latest text-to-image from @GoogleDeepMind.ποΈ Ideogram v3 Quality: latest text-to-image model from @Ideogram_AI. Your votes
0
0
1
Fantastic job from the Apple π.
Apple released DiffuCoder-7B a Masked Diffusion Model for Code Generation. MLX support is a work in progress right now.
0
0
1
Why don't you guys finetune your own datas with good model with THE BEST METHOD @unsloth?! π€. I've made more than 50models with @unsloth and this is one of our best domain models ever π.
You can utilize our Gemma 3n multimodal and fine-tuning Kaggle notebook for any submission to the $150,000 challenge!. The $10,000 is specifically for the Unsloth track - but you can submit it for the main track as well!. Kaggle notebook:
1
0
2
Love this mem0 so much!. But, memvid(repoπ(i'm not a contributor π₯²) is really getting better to memory!. This is still improving!. @Gradio can do many things btw π€
π§ Easily use Mem0 add and search tools with @OpenAI Agents SDK. - add_memory: saves information to memory.- get_memory: retrieves relevant past information. Persistent memory in your agent workflow in just few lines of code. docs below π
1
2
9