• LLMSurgeon: Diagnosing Data Mixture of Large Language Models
• Reasoning with Sampling: Cutting at Decision Points
• CONF-KV: Confidence-Aware KV Cache Eviction with Mixed-Precision Storage for Long-Horizon LLM
• Profiling in PyTorch (Part 1): A Beginner's Guide to torch.profiler
• Harness, Scaffold, and the AI Agent Terms Worth Getting Right
• Grok Build CLI 💻, AI hardware market ⚡, Pope Leo’s AI warning ⛪
• Clipto
• Openstatus MCP Health Checker
• TauricResearch/TradingAgents: TradingAgents: Multi-Agents LLM Financial Trading Framework
• ZhuLinsen/daily_stock_analysis: LLM驱动的 A/H/美股智能分析：多数据源行情 + 实时新闻 + LLM决策仪表盘 + 多渠道推送，零成本定时运行，纯白嫖. LLM-powered stock analysis system for A/H/US markets.
• [D] Monthly Who's Hiring and Who wants to be Hired?
• How Much of a Shortcut Are Connections in Top AI Lab Hiring for PhD grads? [D]

Read full issue →

• Opus 4.8 🧠, Anthropic at $965B 💰, Microsoft’s coding model 👨‍💻
• Import AI 458: Reckoning with the future; and a singularity story
• AI Weekly Issue #497: AI's labor war just went global
• NODUS HN Radar
• Ava Studio
• 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, Cohere, Anthropic, Sagemaker, HuggingFace, VLLM, NVIDIA NIM]
• PatterAI/Patter: Open-source voice-AI SDK. The Vapi/Retell alternative for builders who want to own the stack. Give your AI agent a phone number in 4 lines — Python and TypeScript, MIT licensed, Twilio + Telnyx.
• Notes from the Mistral AI Now Summit
• Show HN: Tiny-vLLM – high performance LLM inference engine in C++ and CUDA

Read full issue →

• Frontier LLM-based agents can overcome the ontology curation bottleneck for natural phenotypes
• Mechanistic origins of catastrophic forgetting: why RL preserves circuits better than SFT?
• MechELK: A Mechanistic Interpretability Framework for Eliciting Latent Knowledge in Large Language Models
• OpenAI private MCP 🤖, Cognition $26B valuation 💰, ElevenLabs Music v2 🎵
• Import AI 458: Reckoning with the future; and a singularity story
• AI Weekly Issue #496: Anthropic's Pentagon model is now everyone's model
• Agent A by Ahrefs
• MCP Bridge by Appfactor
• Netflix/metaflow: Build, Manage and Deploy AI/ML Systems
• OpenHands/OpenHands: 🙌 OpenHands: AI-Driven Development
• Claude Opus 4.8
• YouTube to automatically label AI-generated videos

Read full issue →

• Why LLMs Fail at Causal Discovery and How Interventional Agents Escape
• Agyn: An Open-Source Platform for AI Agents with Scalable On-Demand Execution, Agent Definition as a Code, and Zero-Trust Access
• Tackling Multimodal Learning Challenges with Mixture-of-Expert: A Survey
• Anthropic Microsoft deal 🤝, Cursor $3B ARR 📈, cloud agent lessons 🤖
• AI Weekly Issue #496: Anthropic's Pentagon model is now everyone's model
• AI Weekly Issue #495: Musk, Zuckerberg killed Trump's AI safety order in three phone calls
• Calling Skills for AI Agents
• zero.xyz
• openagents-org/openagents: OpenAgents - AI Agent Networks for Open Collaboration
• HKUDS/LightRAG: [EMNLP2025] "LightRAG: Simple and Fast Retrieval-Augmented Generation"
• I think Anthropic and OpenAI have found product-market fit
• Tech CEOs are apparently suffering from AI psychosis

Read full issue →

• Can LLMs Introspect? A Reality Check
• Is Agent Memory a Database? Rethinking Data Foundations for Long-Term AI Agent Memory
• Your Agents Are Aging Too: Agent Lifespan Engineering for Deployed Systems
• Anthropic Microsoft deal 🤝, Cursor $3B ARR 📈, cloud agent lessons 🤖
• Import AI 458: Reckoning with the future; and a singularity story
• AI Weekly Issue #495: Musk, Zuckerberg killed Trump's AI safety order in three phone calls
• Coworker AI
• Oasis Browser for Mac
• unslothai/unsloth: Unsloth Studio is a web UI for training and running open models like Gemma 4, Qwen3.6, DeepSeek, gpt-oss locally.
• modelscope/evalscope: A streamlined and customizable framework for efficient large model (LLM, VLM, AIGC) evaluation and performance benchmarking.
• I'm Tired of Talking to AI
• Using AI to write better code more slowly

Read full issue →

• Confidence Calibration in Large Language Models
• How Much Thinking is Enough? Quantifying and Understanding Redundancy in LLM Reasoning
• Mixture of Complementary Agents for Robust LLM Ensemble
• Anthropic Microsoft deal 🤝, Cursor $3B ARR 📈, cloud agent lessons 🤖
• Gemini 3.5 Flash ⚡️, Karpathy joins Anthropic 🧑‍💻, OpenAI Guaranteed Capacity ⚡
• AI Weekly Issue #495: Musk, Zuckerberg killed Trump's AI safety order in three phone calls
• Rezonant
• Parsewise API
• av/harbor: Stop configuring your AI stack. Start using it. One command brings a complete pre-wired LLM stack with hundreds of services to explore.
• sgl-project/sglang: SGLang is a high-performance serving framework for large language models and multimodal models.
• Using AI to write better code more slowly
• Norway's 2 petabytes of Huawei flash storage and LLM training

Read full issue →

• AgentWall: A Runtime Safety Layer for Local AI Agents
• Skim: Speculative Execution for Fast and Efficient Web Agents
• The Scaling Laws of Skills in LLM Agent Systems
• Gemini Extended Thinking ✨, ChatGPT finance 📱, Claude Code at scale 👨‍💻
• Claude small business 💼, Anthropic CFO interview 💰, AI adoption data 📊
• Import AI 457: AI stuxnet; cursed Muon optimizer; and positive alignment
• pixserp
• LobeHub
• vllm-project/vllm: A high-throughput and memory-efficient inference and serving engine for LLMs
• reacher-z/ClawBench: Open-source benchmark for browser AI agents on 153 everyday online tasks across 144 live websites. 5-layer recording + DOM-match + LLM judge. Top score 33.3%.
• We stopped AI bot spam in our GitHub repo using Git's –author flag
• Sieve – scans Cursor/Claude chat history for leaked API keys

Read full issue →

• SDOF: Taming the Alignment Tax in Multi-Agent Orchestration with State-Constrained Dispatch
• SkillSmith: Compiling Agent Skills into Boundary-Guided Runtime Interfaces
• TeamTR: Trust-Region Fine-Tuning for Multi-Agent LLM Coordination
• Grok Build 👨‍💻 , Codex customizations 🤖, xAI exodus 👋
• Opus 4.7 Fast ⚡, Qwen Image 2.0 🖼️, serverless GPUs ✨
• Import AI 456: RSI and economic growth; radical optionality for AI regulation; and a neural computer
• Vivago Video Agent
• Loova Agents
• NVIDIA-AI-Blueprints/video-search-and-summarization: Suite of reference architectures for building GPU-accelerated vision agents and AI-powered video analytics applications.
• sipyourdrink-ltd/bernstein: Audit-grade multi-agent orchestration for CLI coding agents (Claude Code, Codex, Gemini CLI, +40 more). HMAC-chained audit log, signed agent cards, per-artefact lineage, air-gap deploy. The orchestrator your compliance team will sign off on. https://bernstein.run
• Most Americans don't trust AI – or the people in charge of it (2025)
• I don't think AI will make your processes go faster

Read full issue →

• Grok Build 👨‍💻 , Codex customizations 🤖, xAI exodus 👋
• Opus 4.7 Fast ⚡, Qwen Image 2.0 🖼️, serverless GPUs ✨
• Nvidia invests $40B 💰, Anthropic acquires compute 🤝, Mistral’s growth 📈
• Loova Agents
• Agentmemory
• sgl-project/sglang: SGLang is a high-performance serving framework for large language models and multimodal models.
• claw-eval/claw-eval: Claw-Eval is an evaluation harness for evaluating LLM as agents. All tasks verified by humans.
• OpenAI and Government of Malta partner to roll out ChatGPT Plus to all citizens
• Frontier AI has broken the open CTF format

Read full issue →

• GraphBit: A Graph-based Agentic Framework for Non-Linear Agent Orchestration
• A Two-Dimensional Framework for AI Agent Design Patterns: Cognitive Function and Execution Topology
• PREPING: Building Agent Memory without Tasks
• Claude small business 💼, Anthropic CFO interview 💰, AI adoption data 📊
• Opus 4.7 Fast ⚡, Qwen Image 2.0 🖼️, serverless GPUs ✨
• Nvidia invests $40B 💰, Anthropic acquires compute 🤝, Mistral’s growth 📈
• TrustClaw by Composio
• Basedash MCP Connectors
• jeremylongshore/claude-code-plugins-plus-skills: 425 plugins, 2,810 skills, 200 agents for Claude Code. Open-source marketplace at tonsofskills.com with the ccpi CLI package manager.
• DemonDamon/AgenticX: AgenticX is a unified, production-ready multi-agent platform — Python SDK + CLI (agx) + Studio server + Machi desktop app. Features Meta-Agent orchestration, 15+ LLM providers, MCP Hub, hierarchical memory, avatar & group chat, skill ecosystem, safety sandbox, and IM gateway (Feishu/WeChat).
• I believe there are entire companies right now under AI psychosis

Read full issue →

• GraphBit: A Graph-based Agentic Framework for Non-Linear Agent Orchestration
• PREPING: Building Agent Memory without Tasks
• EvolveMem:Self-Evolving Memory Architecture via AutoResearch for LLM Agents
• Claude small business 💼, Anthropic CFO interview 💰, AI adoption data 📊
• Opus 4.7 Fast ⚡, Qwen Image 2.0 🖼️, serverless GPUs ✨
• Nvidia invests $40B 💰, Anthropic acquires compute 🤝, Mistral’s growth 📈
• Causo for Fundraising
• Spellar 3.0
• unslothai/unsloth: Unsloth Studio is a web UI for training and running open models like Gemma 4, Qwen3.6, DeepSeek, gpt-oss locally.
• vllm-project/vllm: A high-throughput and memory-efficient inference and serving engine for LLMs
• How Claude Code works in large codebases
• Access to frontier AI will soon be limited by economic and security constraints

Read full issue →

• Think Twice, Act Once: Verifier-Guided Action Selection For Embodied Agents
• Do Androids Dream of Breaking the Game? Systematically Auditing AI Agent Benchmarks with BenchJack
• CHAL: Council of Hierarchical Agentic Language
• Import AI 456: RSI and economic growth; radical optionality for AI regulation; and a neural computer
• Import AI 455: AI systems are about to start building themselves.
• Import AI 453: Breaking AI agents; MirrorCode; and ten views on gradual disempowerment
• CraftBot with Living UI
• Pipali
• langchain-ai/langchain: The agent engineering platform.
• browser-use/browser-use: 🌐 Make websites accessible for AI agents. Automate tasks online with ease.
• Human-level performance via ML was *not* proven impossible with complexity theory [D]
• Trained transformer-based chess models to play like humans (including thinking time) [P]

Read full issue →

• Auto-Rubric as Reward: From Implicit Preferences to Explicit Multimodal Generative Criteria
• SkillLens: Adaptive Multi-Granularity Skill Reuse for Cost-Efficient LLM Agents
• CoCoDA: Co-evolving Compositional DAG for Tool-Augmented Agents
• Import AI 456: RSI and economic growth; radical optionality for AI regulation; and a neural computer
• Import AI 455: AI systems are about to start building themselves.
• Import AI 453: Breaking AI agents; MirrorCode; and ten views on gradual disempowerment
• CraftBot with Living UI
• Crade AI
• langchain-ai/langchain: The agent engineering platform.
• browser-use/browser-use: 🌐 Make websites accessible for AI agents. Automate tasks online with ease.
• TabPFN-3 just released: a pre-trained tabular foundation model for up to 1M rows [R][N]
• Needle: We Distilled Gemini Tool Calling Into a 26M Model

Read full issue →

• Auto-Rubric as Reward: From Implicit Preferences to Explicit Multimodal Generative Criteria
• SkillLens: Adaptive Multi-Granularity Skill Reuse for Cost-Efficient LLM Agents
• CoCoDA: Co-evolving Compositional DAG for Tool-Augmented Agents
• Import AI 456: RSI and economic growth; radical optionality for AI regulation; and a neural computer
• Import AI 455: AI systems are about to start building themselves.
• Import AI 453: Breaking AI agents; MirrorCode; and ten views on gradual disempowerment
• Hyperswitch Prism
• Kelviq
• langchain-ai/langchain: The agent engineering platform.
• browser-use/browser-use: 🌐 Make websites accessible for AI agents. Automate tasks online with ease.
• TabPFN-3 just released: a pre-trained tabular foundation model for up to 1M rows [R][N]
• Needle: We Distilled Gemini Tool Calling Into a 26M Model

Read full issue →