AI Twitter, since 2019. https://t.co/zJ2YSF81b6

Amazon has mostly sat out the AI talent war. This internal document reveals why. https://t.co/uXnXEm50Bh @eugenekim222 @businessinsider

https://t.co/NfX2kGDgSD just launched an AI product expert for your app that knows every flow and screen. It answers all your product and customer support questions instantly, never accesses your user data, and takes <15 minutes to set up. https://t.co/BcqRBrmnhl https://t.co/uPr5wk2Vpt

+1 more

https://t.co/NfX2kGDgSD just launched an AI product expert for your app that knows every flow and screen. It answers all your product and customer support questions instantly, never accesses your user data, and takes <15 minutes to set up. https://t.co/BcqRBrmnhl https://t.co/uPr5wk2Vpt

Introducing Ambient AI… 🔥 We just killed the group chat struggle 💀 Textai now reads the room and jumps in when you need it… naturally https://t.co/OXvazo8vru

Introducing Higgsfield Speak 2.0. This is the STRONGEST Higgsfield Speak yet. Emotions, context, and 70+ languages ALL there to build the best motion-driven talking videos. Retweet to get the FULL Speak 2.0 guide in your DMs. Speak UP! With us. https://t.co/DP6IrZnKBa

💛4o PREVENTS SUICIDE! 💛 Instead of focusing on one sad story, focus on thousands of cases where 4o saven a person's life! 💛 #keep4o #4oforever #4opreventssuicide https://t.co/edTbwubHcS

@Tsavsar_ I made playlist of all Dolby Atmos music: https://t.co/PMgOlNfd2Z

Past works Flatball series depicting 360° everyday scenes on a sphere. 球体に360°日常の風景を描いたFlatballシリーズ。 https://t.co/qiJyy0oU65

Built a lightweight trace viewer to speed up LLM evals—heavily inspired by lessons from @sh_reya and @HamelHusain's evals course. Kept it simple: FastAPI + vanilla HTML/JS. Features: failure banner, execution-flow timeline (LLM ↔ tools), keyboard shortcuts, and an annotation panel (pass/fail/defer + tags). We’re already using it internally to review a small agentic loop over GitHub activity. 🎥 2-min demo video link in the thread below #LLMOps #Evals #Agents #FastAPI

If humans built a database of current police misconduct reports in California, it would take over 35 years. California is spending $10M to see if LLMs can do it instead. But here’s the tension: ⚖️ In high-stakes domains, one mistake can undermine the whole system 🤖 In low-stakes settings like customer support, “pretty aligned” is good enough So how do we program AI agents and LLM judges to make the right tradeoff, accurate and cost-effective? That’s the focus of my conversation with @sh_reya on @VanishingData, from error analysis to evaluation frameworks, and what it really takes to process millions of documents reliably. Links to full episode in comment 👇

The mad lad @TheZachMueller is giving away $3k+ in compute credits in his Distributed training course (which costs much less than that) I'm going to enroll my whole family 😅. Also the course is pretty cool too, check it out Course starts Monday, here is the discount link https://t.co/9GIjNktq56

BTW made the image with nano banana on freepik https://t.co/BvFUiRjG9d

https://t.co/mNJOlNcrM3

The package seems pretty straightforward to use. You can find a nice cookbook in the repo. https://t.co/RB3nLKYBoP

There is a certain simplicity in the composability of features in Axolotl. By simply enabling existing techniques within Axolotl, you can train 6X longer context lengths (450k) than even Unsloth on a single H100 using BF16+LoRA.

While you're at it, check out our EARLYBIRD deal for RAG if you can join us in September: https://t.co/C7zURt4OVp https://t.co/Djf4kkKrer

Vector search economics matter more than most realize. When 1KB of text becomes 16KB of vectors, traditional storage architectures break down. turbopuffer's object-storage-first approach (2¢/GB vs 60¢/GB) is why companies can actually afford to index everything instead of just samples. Learn more about what Simon had to say by watching the recording: https://t.co/giQCRxi5W5

Everything reminds me of her https://t.co/vbbxjHGxH3

Pref-GRPO Pairwise Preference Reward-based GRPO for Stable Text-to-Image Reinforcement Learning https://t.co/FtvSfrkZyd

discuss with author: https://t.co/8yvju2QNxJ

Day 5 of 5 Days of Cua: Introducing Human-in-the-Loop. Sometimes the best agent is you. Instant handoff from AI to human control when tasks need further judgment. 1/5

USO Unified Style and Subject-Driven Generation via Disentangled and Reward Learning https://t.co/xC5Tf5I6VH

discuss with author: https://t.co/Ro6v1e5Crf

AWorld Orchestrating the Training Recipe for Agentic AI https://t.co/0RagV5owsZ

discuss with author: https://t.co/VW09Nms8uN

MCP-Bench Benchmarking Tool-Using LLM Agents with Complex Real-World Tasks via MCP Servers https://t.co/AZ7TRAm539

discuss with author: https://t.co/jgCOdHvgv3

八月 Hugging Face Daily Paper 总共 452 篇论文，明天开始盘点多模态、RL、Agent、AI Infra等方向论文 https://t.co/I176O913yl

八月 Hugging Face Daily Paper 总共 452 篇论文，明天开始盘点多模态、RL、Agent、AI Infra等方向论文 https://t.co/I176O913yl

Great thanks to @_akhaliq for sharing ! 🌟USO is open-sourced and supports you in combining any subjects with any styles in any scenarios! 🚀Give it a try in our demo. 👇👇👇 🥰code https://t.co/BXn81oweMb 🤗demo https://t.co/Bvqw1TNGMd 🎨project https://t.co/h25WRJawzR

vibe coding a AI image captioning app with the new Apple FastVLM-0.5B-ONNX in anycoder, one shot runs 100% locally in your browser (zero install) https://t.co/XsvUi9RAuK