THE INTERNET'S CLEARING HOUSE FOR AI NEWS & INFORMATION

KLUDGE REPORT

Thursday, April 16, 2026  |  08:39 PM UTC
MONITORING 160+ SOURCES  •  UPDATED EVERY 30 MINUTES  •  OPEN SOURCE OR BUST
OPEN SOURCE & BUILDERS
GitHub trending: NousResearch/hermes-agent ⭐596 — The agent that grows with you GitHub trending: safishamsi/graphify ⭐127 — AI coding assistant skill (Claude Code, Codex, GitHub trending: lsdefine/GenericAgent ⭐102 — AI-powered PC agent loop for desktop automat GitHub trending: google/magika ⭐94 — Fast and accurate AI powered file content types detec GitHub trending: alchaincyf/nuwa-skill ⭐75 — 你想蒸馏的下一个员工,何必是同事。蒸馏任何人的思维方式——心智模型、决策启发式、表达DNA GitHub trending: calesthio/OpenMontage ⭐64 — World's first open-source, agentic video prod GitHub trending: yizhiyanhua-ai/fireworks-tech-graph ⭐60 — Claude Code skill for generatin GitHub trending: datawhalechina/hello-agents ⭐57 — 📚 《从零开始构建智能体》——从零开始的智能体原理与实践教程 GitHub trending: luongnv89/claude-howto ⭐53 — A visual, example-driven guide to Claude Cod GitHub trending: KKKKhazix/khazix-skills ⭐49 — 数字生命卡兹克开源的 AI Skills 合集 GitHub trending: sickn33/antigravity-awesome-skills ⭐41 — The Ultimate Collection of 800+ GitHub trending: HKUDS/DeepTutor ⭐39 — "DeepTutor: Agent-Native Personalized Learning Assi GitHub trending: jundot/omlx ⭐37 — LLM inference server with continuous batching & SSD cac GitHub trending: PurpleAILAB/Decepticon ⭐36 — Autonomous Hacking Agent for Red Team Testin GitHub trending: hugohe3/ppt-master ⭐32 — AI generates natively editable PPTX from any doc GitHub trending: ComposioHQ/awesome-claude-skills ⭐32 — A curated list of awesome Claude S GitHub trending: tirth8205/code-review-graph ⭐32 — Local knowledge graph for Claude Code. GitHub trending: github/awesome-copilot ⭐25 — Community-contributed instructions, agents, GitHub trending: teng-lin/notebooklm-py ⭐25 — Unofficial Python API and agentic skill for GitHub trending: usestrix/strix ⭐24 — Open-source AI hackers to find and fix your app’s vu GitHub trending: HKUDS/CLI-Anything ⭐23 — "CLI-Anything: Making ALL Software Agent-Native" GitHub trending: nesquena/hermes-webui ⭐22 — Hermes WebUI: The best way to use Hermes Agen GitHub trending: alirezarezvani/claude-skills ⭐22 — 232+ Claude Code skills & agent plugin GitHub trending: ZhuLinsen/daily_stock_analysis ⭐21 — LLM驱动的 A/H/美股智能分析器:多数据源行情 + 实时新闻 + L
AGENTS & TOOLS
You opened Claude, got something generic, and closed the tab. Here’s what was missing. Qwen3.6-35B-A3B on my laptop drew me a better pelican than Claude Opus 4.7 datasette.io news preview datasette-export-database 0.3a1 datasette 1.0a27 Quoting John Gruber Gemini 3.1 Flash TTS Gemini 3.1 Flash TTS Quoting Kyle Kingsbury datasette-ports 0.3 Zig 0.16.0 release notes: "Juicy Main" datasette PR #2689: Replace token-based CSRF with Sec-Fetch-Site header protection Trusted access for the next era of cyber defense Current AIs seem pretty misaligned to me Do not conquer what you cannot defend Specialization is a Driver of Natural Ontology You Aren't in Charge of the Overton Window; Politics Is Not Interior Design The Mirror Test Is Complicated Serverless GPU Inference Cost Comparison: Roboflow, GCP, AWS, Azure How to Use Roboflow to Create Conditional Annotators
PAPERS & RESEARCH
Cross-Layer Co-Optimized LSTM Accelerator for Real-Time Gait Analysis SFT-GRPO Data Overlap as a Post-Training Hyperparameter for Autoformalization Artificial intelligence application in lymphoma diagnosis with Vision Transformer using we A Study of Failure Modes in Two-Stage Human-Object Interaction Detection Cognitive Offloading in Agile Teams: How Artificial Intelligence Reshapes Risk Assessment Jump-Start Reinforcement Learning with Vision-Language-Action Regularization Med-CAM: Minimal Evidence for Explaining Medical Decision Making Free Geometry: Refining 3D Reconstruction from Longer Versions of Itself Towards Fine-grained Temporal Perception: Post-Training Large Audio-Language Models with A ESCAPE: Episodic Spatial Memory and Adaptive Execution Policy for Long-Horizon Mobile Mani TokenFormer: Unify the Multi-Field and Sequential Recommendation Worlds MaMe & MaRe: Matrix-Based Token Merging and Restoration for Efficient Visual Perception an Beyond Conservative Automated Driving in Multi-Agent Scenarios via Coupled Model Predictiv ROSE: Retrieval-Oriented Segmentation Enhancement RobotPan: A 360$^\circ$ Surround-View Robotic Vision System for Embodied Perception A 3D SAM-Based Progressive Prompting Framework for Multi-Task Segmentation of Radiotherapy
BIG AI WATCH
Codex for (almost) everything Introducing GPT-Rosalind for life sciences research Accelerating the cyber defense ecosystem that protects us all The next evolution of the Agents SDK Gemini 3.1 Flash TTS: the next generation of expressive AI speech How to make remarkable videos with Seedance 2.0 Early Indicators of Reward Hacking via Reasoning Interpolation Introducing Claude Opus 4.7 The PR you would have opened yourself Training and Finetuning Multimodal Embedding & Reranker Models with Sentence Transformers Inside VAKRA: Reasoning, Tool Use, and Failure Modes of Agents Meet HoloTab by HCompany. Your AI browser companion. Designing synthetic datasets for the real world: Mechanism design and reasoning from first AI-generated synthetic neurons speed up brain mapping OpenAI turns Codex into an always-on coding agent that watches your screen OpenAI says more women than men now use ChatGPT, flipping an 80-20 male split at launch Nvidia wants to scale robot simulation training with Lyra 2.0 India's 1.5 million annual IT graduates face an industry that's moving on without them Google launches native Gemini app for Mac Anthropic's Claude Opus 4.7 makes a big leap in coding, while deliberately scaling back cy
CRITICS & SKEPTICS
Peak absurdity, Part II Commerce Secretary Howard Lutnick’s old firm pumps $10 million into super PAC led by Tethe Pluralistic: A Pascal's Wager for AI Doomers (16 Apr 2026) Pluralistic: Rights for robots (15 Apr 2026) App Stores Push Users Toward Nudify Apps, New Research Shows I Almost Lost My Mind in the Bridal Algorithm Thomson Reuters Shareholders Demand Investigation into ICE Contracts Ukraine Says Russians are Surrendering to Robots Emails Reveal Space Force’s Hardest Mission Is Writing a Song Podcast: How the FBI Extracted Deleted Signal Messages The Mexican security company with a $1.27 billion surveillance empire Voice actors fight to save their livelihoods and local cultures from Hollywood’s AI push Open-world evaluations for measuring frontier AI capabilities Stop New York's Attack on 3D Printing How Push Notifications Can Betray Your Privacy (and What to Do About It) EFF Calls on Kuwait to Release Journalist Ahmed Shihab-Eldin
PREDICTION MARKETS
[40%] Will the DC Circuit rule for Anthropic on the merits in the Pentagon supply chain ca [35%] Will the US lift its naval blockade of the Strait of Hormuz by May 31, 2026? [78%] Will OpenAI publicly release the model codenamed 'Spud' (as GPT-5.5, GPT-6, or any o [50%] Daily Coinflip [51%] Daily Coinflip [34%] Will Anthropic's lawsuit against the Pentagon be settled out of court before August [50%] Daily Coinflip [46%] On June 1, will OpenAI be valued higher than Anthropic? [63%] Will Anthropic hit a trillion before OpenAI gets back to a trillion? Which Claude / Anthropic model(s) will be released today? (April 16 - Flash Market)
COMMUNITY BUZZ
Qwen3.6-35B-A3B released! Claude Opus 4.7 benchmarks Anyone else get more excited for new open source models than new flagship ones? Video of how my LLM's decoder blocks changed while training Released Qwen3.6-35B-A3B Opus 4.7 has been spotted on Google Vertex I’ll take an open-model release over a closed SOTA any day, who’s with me? Opus 4.7 seems to rolled out to Claude Web Figure.AI new balance policy allows their 03 humanoid robot to keep its balance even if so More reasons to go local: Claude is beginning to require identity verification, including Claude Opus 4.7 My fresh experience with the new Qwen 3.6 35B A3B started on a long note. White House Moves to Give US Agencies Anthropic Mythos Access Failure to Reproduce Modern Paper Claims [D] AI Is Weaponizing Your Own Biases Against You: New Research from MIT & Stanford I got it guys, I think I finally understand why you hate censored models
LEADERBOARDS & BENCHMARKS
HF Model trending: google/gemma-4-31B-it (1,972 likes, 3,195,626 downloads) HF Space trending: mrfakename/Z-Image-Turbo (2925 likes) HF Space trending: multimodalart/qwen-image-multiple-angles-3d-camera (2252 likes) HF Space trending: r3gm/wan2-2-fp8da-aoti-preview (2065 likes) HF Dataset trending: nvidia/PhysicalAI-Autonomous-Vehicles (836 likes) HF Space trending: selfit-camera/Omni-Image-Editor (1424 likes) HF Space trending: microsoft/TRELLIS.2 (1406 likes) HF Model trending: zai-org/GLM-5.1 (1,277 likes, 94,376 downloads) HF Model trending: dealignai/Gemma-4-31B-JANG_4M-CRACK (1,154 likes, 143,000 downloads) HF Model trending: k2-fsa/OmniVoice (596 likes, 701,830 downloads) HF Model trending: MiniMaxAI/MiniMax-M2.7 (842 likes, 142,955 downloads) HF Model trending: openbmb/VoxCPM2 (938 likes, 15,249 downloads)
RADIO / PODS
▶ Uber, Nissan, and Mercedes Chose This Self-Driving Startup | Alex Kendall, Wayve ▶ My bets on open models, mid-2026 ▶ Open Source Self-Driving with Comma AI ▶ AI's Great Divergence ▶ Vibe Coding Gets an Upgrade ▶ AI Populism Turns Violent ▶ We're All Building a Single Digital Assistant ▶ Notion’s Token Town: 5 Rebuilds, 100+ Tools, MCP vs CLIs and the Software Factory Future — ▶ #334 Abhishek Singh: The $1.2 Billion Plan to Turn India Into an AI Superpower ▶ #333 Adi Kuruganti: Why Your AI Pilot Is Failing and What It Takes to Reach Production ▶ Technology, Culture, and the Next AI Interface with signüll ▶ Replit's CEO on Vibe Coding, Wealth Building, and What Most People Get Wrong About AI