News
AI News Brief — May 15
GitHub velocity is led by langgenius/dify; paper attention is clustering around Edit-Compass & EditReward-Compass: A Unified Benchmark for Image Editing and Reward Modeling; social attention is tilting toward Lessons learned from inside China’s AI lab; biggest mover: MemPalace/mempalace (+2). 10 repo signals, 10 paper picks, and 10 community items made today's cut.
GitHub velocity is led by langgenius/dify; paper attention is clustering around Edit-Compass & EditReward-Compass: A Unified Benchmark for Image Editing and Reward Modeling; social attention is tilting toward Lessons learned from inside China’s AI lab; biggest mover: MemPalace/mempalace (+2). 10 repo signals, 10 paper picks, and 10 community items made today's cut. This fallback brief keeps the page live when the optional Gemma pass times out.
Signal Board
Repositories and papers
Keep the full repo and paper scan above the fold, then read the day as one short brief below.
Top list
Repository Momentum
Fresh GitHub projects worth scanning before the feed turns over.
langgenius/dify
Production-ready platform for agentic workflow development. Updated 1h ago. 141370 stars, +800/7d, created 1128d ago.
anomalyco/opencode
The open source coding agent. Updated 1h ago. 160198 stars, +800/7d, created 379d ago.
code-yeongyu/oh-my-openagent
omo; the best agent harness - previously oh-my-opencode. Updated 1d ago. 57597 stars, +800/7d, created 163d ago.
paperclipai/paperclip
The open-source app everyone uses to manage agents at work. Updated 23h ago. 65067 stars, +800/7d, created 73d ago. Down 1 spots from the previous run.
rtk-ai/rtk
CLI proxy that reduces LLM token consumption by 60-90% on common dev commands. Single Rust binary, zero dependencies. Updated 23h ago. 47281 stars, +800/7d, created 112d ago. Up 1 spots fro…
ZhuLinsen/daily_stock_analysis
LLM-powered stock analysis system for A/H/US markets: multi-data source market + real-time news + LLM decision-making dashboard + multi-channel push, zero-cost scheduled operation, pure fre…
BerriAI/litellm
Python SDK, Proxy Server (AI Gateway) to call 100+ LLM APIs in OpenAI (or native) format, with cost tracking, guardrails, loadbalancing and logging. [Bedrock, Azure, OpenAI, VertexAI, Coher…
Skyvern-AI/skyvern
Automate browser based workflows with AI. Updated <1h ago. 21606 stars, avg 26.8/day, created 806d ago.
MemPalace/mempalace
The best-benchmarked open-source AI memory system. And it's free. Updated 1d ago. 52127 stars, +800/7d, created 40d ago. Up 2 spots from the previous run.
antirez/ds4
DeepSeek 4 Flash local inference engine for Metal and CUDA. Updated 1h ago. 8688 stars, created 8d ago. Down 7 spots from the previous run.
Top list
Fresh Papers
New research worth bookmarking for a deeper read.
Edit-Compass & EditReward-Compass: A Unified Benchmark for Image Editing and Reward Modeling
Recent image editing models have achieved remarkable progress in instruction following, multimodal understanding, and complex visual editing. However, existing benchmarks often fail to fait…
Learning Agentic Policy from Action Guidance
Agentic reinforcement learning for large language models leverages action data from human interactions as reference guidance to improve exploration and reduce dependence on costly supervise…
HAGE: Harnessing Agentic Memory via RL-Driven Weighted Graph Evolution
HAGE introduces a weighted multi-relational memory framework that enables query-conditioned traversal over unified relational memory graphs, improving long-horizon reasoning accuracy throug…
PresentAgent-2: Towards Generalist Multimodal Presentation Agents
PresentAgent-2 is an agentic framework that generates presentation videos from user queries by conducting research, creating multimodal slides, and producing interactive content across sing…
Qwen-Image-VAE-2.0 Technical Report
Qwen-Image-VAE-2.0 is a high-compression Variational Autoencoder suite that improves reconstruction fidelity and diffusability through enhanced architecture, large-scale training, and seman…
TrackCraft3R: Repurposing Video Diffusion Transformers for Dense 3D Tracking
TrackCraft3R enables efficient dense 3D tracking from monocular video by adapting video diffusion transformers to follow physical points across frames using dual-latent representation and t…
MemReread: Enhancing Agentic Long-Context Reasoning via Memory-Guided Rereading
MemReread addresses long-context reasoning challenges by avoiding intermediate retrieval and employing question decomposition with rereading to recover discarded information, maintaining li…
MinT: Managed Infrastructure for Training and Serving Millions of LLMs
MinT is a managed infrastructure system that enables efficient low-rank adaptation training and serving by keeping base models resident and moving lightweight adapter revisions, scaling acr…
Retrieval is Cheap, Show Me the Code: Executable Multi-Hop Reasoning for Retrieval-Augmented Ge…
Multi-hop question answering is reformulated as program synthesis and execution, enabling structured reasoning, deterministic feedback, and improved performance over existing retrieval-augm…
FrameSkip: Learning from Fewer but More Informative Frames in VLA Training
FrameSkip is a data-layer frame selection method that improves VLA policy training by prioritizing high-importance frames based on action variation and visual-coherence metrics. Surfaced vi…
Today in AI
The day in one pass
GitHub velocity is led by langgenius/dify; paper attention is clustering around Edit-Compass & EditReward-Compass: A Unified Benchmark for Image Editing and Reward Modeling; social attention is tilting toward Lessons learned from inside China’s AI lab; biggest mover: MemPalace/mempalace (+2). 10 repo signals, 10 paper picks, and 10 community items made today's cut.
The quickest scan starts with langgenius/dify, Edit-Compass & EditReward-Compass: A Unified Benchmark for Image Editing and Reward Modeling, and Skyvern-AI/skyvern, while 10 GitHub-led signals anchor the repo side of the brief.
Use the structured digest below when you need the full wire, raw links, and the longer tail of items that did not make the front narrative.
Archive
Recent issues
Generated from the ranked feed for May 15, 2026 as one single daily issue.
Linked Mentions
No linked mentions yet.