Sitemap
75 articlesA complete index of every article on this site.
AI
- What a New Mixture-of-Experts Routing Application Claims, and Where It Lands in G06N
- Autodesk's Just-Issued US 12,657,468 Trains Graph Neural Nets to Read 3D CAD Geometry
- NVIDIA's Just-Issued US 12,657,819 Trains a Road-Surface DNN on Math, Not Miles
- WorkflowView Turns Noisy Interaction Logs Into Workflows an LLM Can Read
- SIMMER Catches the Planning Failures That Don't Stop the Plan
- GitOfThoughts Makes Agent Reasoning Replayable - and Then Admits Memory Rarely Helps
- A Production Agent's Worst Failure Isn't Crashing - It's Lying Convincingly
- AdaSR Teaches Reasoning Models When to Think While the Input Is Still Arriving
- Parallel-Synthesis Lets an Agent Read Its Workers' KV Caches Instead of Their Text
- Just Issued: NVIDIA's Grant on Synthesizing Video From Audio (G06V, G10L)
- Just Issued: NVIDIA's Grant on Data-Set Generation and Augmentation (G06V)
- Just Published: Google's Application on a Mixture-of-Expert Speech Model (Published, Not Granted)
- IP Strategy Brief: Reading Anthropic's Computer-Use Patent Family as a Competitive Instrument
- Claim Breakdown: The Independent Claim in Anthropic's Computer-Use Agent Patent
- Whitespace Watch: Where the Inference-Efficiency Patents Are Clustering — and Where They Aren't
- Landscape Report: The Vision-Language Patent Cluster Across Salesforce, Microsoft, and Google
- Just Issued: Microsoft's Grant on Routing Models by Perplexity — What Claim 1 Covers (G06N)
- Who Owns Mixture-of-Experts? Tracing the Foundational Claims Through PatentBear
- What CPC Class Are AI Patents Filed Under? A Primer on G06N, G06F, and G06V
- Granted vs. Published: What the Difference Actually Means for an AI Patent
- Just Published: An OpenAI-Linked Application on Optimizing the KV-Cache for LLM Inference (2025)
- Just Published: NVIDIA's Application on Foundation Models for Multimodal Data Selection (2025)
- Claim Breakdown: Google's Patent on Fine-Tuning LLMs with Search-Engine Feedback
- Disclosure Cross-Check: The 2025 KV-Cache Patent Race and the AI Serving-Cost Story
- Just Issued: HPE's 2025 Grant on Efficient KV-Cache Management for LLMs
- Just Published: Intel's Application on Dynamic KV-Cache Quantization for LLM Serving (2025)
- Just Published: Baidu's Application on a Mixture-of-Experts Implementation (2025)
- Just Published: An Application on Full-Stack Hardware-Accelerator Search (2024)
- Landscape Report: The 2023-24 Text-to-Image Diffusion Patent Surge
- Just Issued: A 2024 Grant on Multi-Algorithm Diffusion Sampling for Text-to-Image
- Who Owns Mixture-of-Experts? The 2024 Update
- Just Issued: Intel's 2024 Grant on a Sparse Matrix-Multiplication Accelerator
- Just Published: Microsoft's Application on a Mixture-of-Experts Layer With Dynamic Gating (2024)
- Just Issued: Google's 2024 Grant on Performing Matrix Multiplication in Hardware
- Just Issued: Salesforce's 2023 Grant on a Generative Few-Shot Sentiment Model
- Just Issued: Intuit's 2023 Grant on Context-Biased AI Video Generation
- Just Issued: IBM's 2023 Grant on BERT-Based FAQ and Document Retrieval
- Just Published: Microsoft's Application on Synthetic Data for Training NLU Models (2023)
- Just Issued: Salesforce's 2023 Grant on a Unified Vision-and-Dialogue Transformer
- Who Owns Federated Learning? A 2022 Ownership Map
- Just Issued: Microsoft's 2022 Grant on a Generalized Reinforcement-Learning Agent
- Just Issued: Amazon's 2022 Grant on RL-Trained Compression Policies for Models
- Just Issued: NVIDIA's 2022 Grant on Neural Rendering for Inverse Graphics
- Just Published: Intel's Application on Dynamic Precision for Integer Deep Learning (2022)
- Just Issued: NVIDIA's 2022 Grant on Simultaneous Compute and Graphics Scheduling
- Granted vs. Filed: How to Read the 2021 AI Patent Record Without Overstating It
- Just Published: Microsoft's Application on Disentangled-Attention Transformers (2021)
- Whitespace Watch: The Federated-Learning Privacy Filings of 2021
- Just Published: NVIDIA's Application on Fully-Fused Neural-Network Execution (2021)
- Just Issued: Intel's 2021 Grant on Per-Kernel K-Means Compression for Networks
- Just Issued: Baidu's 2021 Grant on Anchor Boxes for Self-Driving Object Detection
- Just Published: Intel's Application on Loss-Aware Low-Bit Quantization (2021)
- Landscape Report: Who Was Filing Neural-Network Training Patents in 2020
- Just Issued: Google's 2020 Grant on Speech Recognition With Recurrent Networks
- Just Issued: Microsoft's 2020 Grant on Controllable Conditional Image Generation
- Just Issued: Adobe's 2020 Grant on Image Composites Using a GAN
- Just Issued: IBM's 2020 Grant on Cascaded Neural Networks That Train Each Other
- Just Issued: Salesforce's 2020 Grant on End-to-End Speech Recognition With Policy Learning
- Just Issued: Google's 2020 Grant on Training a Document-Classification Network
AI IP
AI Patents
- Who Owns the Patents on Transformers and Attention? Reading the G06N Landscape Through Real Grants
- Claiming a Neural Network: How 35 U.S.C. 112 Shapes Machine-Learning Patent Claims
- Can an AI Be a Named Inventor on a Patent? What Thaler v. Vidal and USPTO Guidance Establish
- What CPC Class Are AI Patents Filed Under? A Primer on G06N and the Machine-Learning Subgroups
- AI Patent Subject-Matter Eligibility: How the Alice/Mayo Test Decides Whether a Machine-Learning Claim Survives 35 U.S.C. 101
Artificial Intelligence
- A Newly Issued Patent Claims Distilling a Server Speech Model Into On-Device Encoders Without Moving Raw Audio
- New Off-Policy Evaluation Method Targets Rewards That Go Missing Not at Random
- Percolation Clusters Proposed as a Structured Testbed for Interpretability Methods
- Decoupling the Task From the Robot: A Meta-RL Framework That Reuses Knowledge Across Bodies
- EvolveNav Gives Navigation Agents a Rule Memory That Rewrites Itself at Test Time
- Ternary Mamba Cuts the Token Bill for 1.58-Bit State Space Models by 1,000x
- A PDE Replaces Score Matching: Kolmogorov Regression Targets the Drift in Diffusion Policies
- Fixed-Point Reasoners Use Convergence Itself as a Transformer's Stop Signal
- Looped World Models Claim 100x Parameter Efficiency by Recycling One Transformer Block