25/04/26·25/04/26AIDeepSeek V4: 1M‑Token Context and Budget Frontier AI ModelsDeepSeek V4 launches with 1M‑token context, MoE‑based Pro and Flash models, and prices far below U.S. rivals, focusing o…DeepSeekAlibabaOpen Source
25/04/26SecurityPalantir Manifesto Hits at Regressive Cultures and AI ShiftThe backlash comes from Palantir framing some cultures as “regressive and harmful” while arguing tech firms must help de…Palantir
22/04/26AIOpenAI Updates ChatGPT Images With Better TextOpenAI’s update makes image generation more reliable for posters, slides, charts, and structured visuals.OpenAIChatGPT
20/04/26AINews Giants Block Wayback Machine Over AI FearsBig news sites block Internet Archive crawlers to stop AI scraping via Wayback Machine, risking erasure of web history.
19/04/26AIAnthropic Launches Claude Design to Rival Figma ToolsNew Claude tool builds prototypes, slides from chat. Reads codebases for consistency, exports to Canva/PPTX. Speeds up d…AnthropicClaude
18/04/26AIExploring Qwen3.6: Coding Benchmarks and SpeedQwen3.6-35B-A3B: Open MoE LLM with 35B params (3B active), tops SWE-bench at 73.4%. Fast local runs at 170 t/s, strong a…QwenAlibabaOpen Source
12/04/26·13/04/26SecurityEurope's Palantir Boom Amid Sovereignty and Rights FearsPalantir stock falls amid 2026 jitters, but EU investors pour $27B despite rights scandals with ICE, Israel. Sovereignty…Palantir
12/04/26AIClaude Mythos Leak Ignites Fears of Unstoppable AI ExploitsAnthropic's Mythos tops Opus 4.6 by 16.5% in cyber benchmarks, uncovers old flaws; private preview via Google Cloud. AnthropicClaude
08/04/26AIChinese AI Models Dominate OpenRouter Top Six in Token UsageChinese LLMs took top 6 on OpenRouter with 13T tokens vs US 3T; Qwen3.6 Plus led at 4.6T. Five weeks straight ahead. ClaudeAlibabaQwen+3
08/04/26AIInside the Claude Code Leak and Anthropic’s Agent DesignAnthropic’s coding agent leak shared 500k lines of TypeScript, agent loops, and context compression, yet kept Claude’s m…ClaudeAnthropic
06/04/26AINVIDIA’s AI Chip Share in China Drops from 95% to 55%NVIDIA’s AI chip share in China has fallen from about 95% to 55% as domestic accelerators gain ground.NVIDIAHuawei
30/03/26AIGoogle’s TurboQuant makes AI caches smaller and fasterGoogle's TurboQuant compresses LLM KV cache to 3 bits, slashing memory 6x and speeding attention 8x on H100s with zero a…Google
28/03/26·28/03/26AIFLUX.2 klein 9B-KV Explained: Speed, Quality, GPUsFLUX.2 klein 9B-KV is a fast image model for editing and repeated references, with up to 2.5x faster multi-reference wor…FluxComfyUI
28/03/26AIKVTC: Nvidia’s 20x LLM Memory Cut Without RetrainingNvidia’s KVTC compresses LLM KV caches up to 20x, cutting GPU memory and speeding responses without changing model weigh…NVIDIA
27/03/26AISora’s Short Life: Inside OpenAI’s Quiet RetreatThe Sora shutdown shows that running a consumer deepfake engine is costly, risky and strategically messy for OpenAI.OpenAI
21/03/26AIGoogle Stitch: From simple prompt to working app UIGoogle Stitch turns prompts and sketches into multi‑screen web and mobile UIs, ready to export to Figma or HTML/CSS for …GoogleGemini
11/03/26AIYann LeCun’s AMI Lab Pioneers Physical‑World AITuring winner Yann LeCun launches AMI to prove that world‑understanding AI can beat today’s chatbot‑style language model…
10/03/26SecurityClaude, Palantir and Who Controls AI in Modern WarAnthropic’s clash with the Pentagon shows how hard it is to keep AI out of autonomous weapons and mass surveillance on t…AnthropicClaude
06/03/26·06/03/26AIOpenSandbox: A Unified Sandbox Layer For AI AgentsAlibaba’s OpenSandbox offers a unified, secure API so AI agents can run code, browse and train in isolated Docker or Kub…AlibabaOpen Source
06/03/26·06/03/26AIAlibaba's Tiny Qwen Beats Big OpenAI ModelAlibaba's Qwen3.5-9B tops OpenAI's gpt-oss-120B on GPQA, Video-MME benchmarks. Runs on laptops, multimodal edge AI star.AlibabaQwenOpen Source
