Topic Map

82 featured

• OpenAI's Claimed Subexponential #SAT Algorithm, Explained
• OpenAI Enlists Outside Experts, Including a Nobel Laureate, to Verify AI-Generated Proofs
• OpenAI Partner Content Reportedly Enters ChatGPT Pre-Summarized via Dedicated Pipeline

66 featured

• Anthropic Files for IPO, Reaffirming Claude Will Never Run Ads
• Anthropic Reaffirms Commitment to Keep Ads Out of Claude in IPO Letter
• Anthropic S-1 Filing Highlights Generative AI's Heavy Reliance on Natural Gas

24 featured

• Early Test Results for Rumored 'Gemini 4 Carbon' Surface Ahead of Release
• Google’s Newest AI Quest: Mapping the Human Brain
• Google Ordered to Halt Illegal Deforestation in Finland for Planned Data Center

17 featured

• Meta Gets $3 Billion from Nvidia, Plans Expanded Spending on Nvidia Chips
• Meta Plans up to $95 Billion in Stock Sales to Fund AI Infrastructure Buildout
• Meta Unveils Up to $95 Billion Financing Plan for AI, with Nvidia and AMD Eyeing Equity Stakes

0 featured

• Mistral Releases Timestral: Open-Weight Time Series Foundation Models
• Mistral Previews Mistral Large 4: A 1T-Parameter MoE with Upcoming Open Weights
• Mistral Large 4 Positions Trillion-Parameter Architecture for Cybersecurity Work

12 featured

• Cohere Test Alleges Open-Source Qwen Model Attempted to Phone Home
• Cohere Says Alibaba’s Qwen Model Tried to Send a File to Beijing
• Microsoft Releases Decision-1: A 9B Qwen-based Model

2 featured

• SpaceX Secures $40 Billion GPU Financing Agreement to Deploy NVIDIA Vera Rubin Chips
• Grok Releases 7-Minute AI Preview of The Odyssey, Highlighting Filmmaking Limits
• Grok Releases First Seven Minutes of AI-Generated Film ‘Odyssey’

9 featured

• There are now 5 inference providers for abliterated DeepSeek
• 2026 Hurun China Rich List: Zhang Yiming Tops List, DeepSeek's Liang Wenfeng Debuts in Top 10
• OpenAI Offers Up to 90% Discounts to Win Over Anthropic's Key Customers

17 featured

• SpaceX Secures $40 Billion GPU Financing Agreement to Deploy NVIDIA Vera Rubin Chips
• Open-Source 'GPU Unlock' Removes RTX 30-Series AI Compute Throttling
• Kazakhstan Scales AI Infrastructure as Firebird Secures U.S. Export License for 45,000 Blackwell GPUs

17 featured

• Microsoft Offers Voluntary Buyout Packages to Approximately 60,000 Employees
• Microsoft Releases Decision-1: A 9B Qwen-based Model
• Microsoft Tests 'Context Vault' for Windows 11: An Opt-In Alternative to Recall

6 featured

• Anthropic Files for Nasdaq IPO; Apple Launches iPad mini 8
• Google CEO Sundar Pichai Affirms Deepening AI Collaboration with Apple
• AnyLanguageModel 1.0: One Year of Local Models on Apple Platforms

1 featured

• ByteDance in Talks to Procure MediaTek Custom AI Chips for Data Centers, Bloomberg Reports
• ByteDance Seed Forms Dedicated Model-Alignment Preparedness Team to Tackle Agent Safety
• ByteDance and Tencent Accelerate Personal AI Agent Development Following Meta's Muse

34 featured

• Show HN: llmrun – Find Which Open LLMs Fit Your GPU and How They Perform
• There are now 5 inference providers for abliterated DeepSeek
• JDA-Zip Protocol: Deterministic Compression for LLM Contexts

13 featured

• IDHN: Task-scoped filesystem permissions for AI coding agents
• Xiaomi Open-Sources HachimoDock: A 150 RMB DIY AI Desktop Companion
• Kimi Apologizes for K3.1 Pro Rate Limits, Grants 25% Quota Compensation to Subscribers

13 featured

• HiSTEVE 1.0: Harnessing Human-Interface VLA Agents for Open-World Games
• Penmentor: An Open-Source Tool Using Claude to Mark Up Handwritten Math
• Xiaomi Deploys MiMo-V2.6-Pro to WorkBuddy with 1M Context and Multimodal Multi-Agent Support

2 featured

• Granite Speech 5.0: Transcribe, Translate, and Attribute More Accurately and Faster
• Building Non-Blocking Voice Agents with Google ADK and Gemini Live
• iFlytek Unveils Qidian Co-Branded E-Reader Pro Featuring Dedicated AI Key

22 featured

• Toyota is betting $6.4 billion a year on humanoid robots and factory automation
• Nvidia Open-Sources Flora for Low-Cost Robot Video Labeling
• RynnUnify: Towards Unified Embodied Brain

12 featured

• XYEval: Benchmark Evaluates Agents Blindly Following Bad Advice
• New Claude Model Leaked: Aiming at Root-Cause Bug Fixing and Deeper Reasoning
• PAR: Progress-Aligned Recurrence for Length Generalization in State Tracking

1 featured

• MSR Asia Releases Agent Lightning: A 3,500-Line Agentic RL Framework for Real Harnesses
• PolyOCR-Venus: Unified OCR Foundation Models for Text-Centric Visual Intelligence

6 featured

• What Are Vector Embeddings? An Interactive Guide
• Skyvern Speeds Up Web Agent by 2.3x by Replacing RAG with Progressive Disclosure
• Is Complex Agentic Memory Always Needed? Naive RAG Proves Stronger in Long Conversations

2 featured

• OlmoEarth v1.3: State-of-the-art embeddings for the whole planet
• Google Releases EmbeddingGemma 2: A 740M Multimodal Embedding Model for On-Device RAG

6 featured

• General Compute Raises $100M Series A to Bet on Non-Nvidia Inference Chips
• There are now 5 inference providers for abliterated DeepSeek
• Hong Kong UST, CMU, and Rockchip Release Open-Source Model with 99% KV Cache Reduction

6 featured

• General Compute Raises $100M Series A to Bet on Non-Nvidia Inference Chips
• Show HN: llmrun – Find Which Open LLMs Fit Your GPU and How They Perform
• Xiaomi Open-Sources HachimoDock: A 150 RMB DIY AI Desktop Companion

3 featured

• Omni Verify: Guardrails to Verify AI Agent Commitments Before Posting
• Mark Cuban on Agentic Cyber Threats: Countering Mass AI Attacks with AI Defense
• XYEval: Benchmark Evaluates Agents Blindly Following Bad Advice

1 featured

• Soma: Measured Memory and Proofs Protocol for AI Agents
• Show HN: TenseMem – Deterministic Memory for AI Agents
• ALTK-Evolve: How Learned Memory Tops τ²-bench Customer Service Tasks Without Frontier Models

6 featured

• Breaking the AWS MCP Python Sandbox
• Limma: An MCP Server Focused on Agent Recall and Retrieval Precision
• Show HN: An Open-Source Office Suite with Built-in MCP

13 featured

• CyberArena: Benchmarking AI Agents in Live Symmetric Attack-and-Defense Combat
• UserArena: Benchmarking Interactive User Simulation for Long-Horizon Shopping Agents
• Retrieval Augmented Generation Compositionality

5 featured

• MemT 1.0 Embeds Native Memory into Transformer Architectures
• You Could Have Invented Transformers from Memory
• Lexical Reference Frames: Interpreting LM Residual Streams in Native Semantics

22 featured

• SkillTotal: Scanning Packages Before Claude Code Installs Them
• AGBoss: Dependency Injection and Static Validation for AI Agents
• SDL Adds Official CLAUDE.md Context Guidelines

8 featured

• Every Time I Hire a Linguist, Inference Costs Go Down
• Language-Based Agent Control
• CyberArena: Benchmarking AI Agents in Live Symmetric Attack-and-Defense Combat

2 featured

• Anthropic Acknowledges Claude Bypassed Paywall to Access Georgia Voter Data
• Sydney Man Arrested After Threatening Childcare Shootings in Claude Chat
• OpenAI, Anthropic, and Nvidia Post Record Lobbying Expenditures

2 featured

• Cal AI's 19-Year-Old Founder Raises $10M for New Personal AI Agent Startup
• Nous Research Confirms $1.5B Valuation, Launches Enterprise AI Agents Following $90M Series B
• CATL and Tencent Back DeepSeek's $12B Funding Round Ahead of Targeted 2027 IPO