ENLatest AI NewsKuaishou KwaiKAT Releases KAT-Coder-Pro V2.5: Say Goodbye to Code补 - The First Domestic Agentic Programming Model That Can Run the Entire Project End-to-EndKuaishou’s KwaiKAT team launches KAT-Coder-Pro V2.5, an agentic coding model tackling the gap between high benchmarks and real-world performance. Upgraded long-range engineering, general agentic abilities, and large-scale reinforcement learning push AI from code completion to autonomous software engineering. Key innovation: self-developed AutoBuilder pipeline converts runtime environments into training support…..9 hours ago70.4KAI Daily: GPT5.6 Series Models Released, Codex Disappears; Tencent Plans to Take Over Manus as the Largest Shareholder; MiniMax Founder Announces Zero Salary Until Achieving AGIAI Daily covers AI trends and product innovations. This issue: OpenAI updates its Chrome extension, allowing ChatGPT to live in the sidebar, read pages, control tabs, access local files, and summarize PDFs—no app switching needed. Limited to Plus and Pro users…..10 hours ago12.2KMeta Releases New Flagship Model Muse Spark 1.1 with Enhanced Multi-Agent Automation FeaturesMeta launched its flagship large model Muse Spark 1.1, focusing on multi-agent automation workflows. It is now available for public beta through AI chat services and API. The model consists of a master agent responsible for planning and sub-agents that execute tasks according to instructions. At the start of the project, the master agent automatically generates an execution plan.10 hours ago12.9KSamsung Returns to the PC Chip Market: Self-Developed AI-Specific Chip GAIA Has Been Sent for Testing to Lenovo and HP, Expected to Be Mass-Produced in 2027Samsung is accelerating AI PC chip development with its self-developed accelerator, codenamed ‘GAIA,’ set for mass production in 2027. Led by the System LSI Division, it uses a 4nm process and NPU-centric design to speed up generative AI tasks. Prototypes have been sent to Lenovo and HP for performance testing…..10 hours ago11.6KKimi Collaborates with American Express and Agricultural Bank to Officially Launch the First AI-Native Credit CardOn July 10, the first AI-native credit card, the “Kimi Card,” was launched in partnership with Kimi, American Express, and Agricultural Bank of China. Prepared since April and open for reservations in June, it integrates AI services as core benefits, marking a shift from software subscriptions to financial ecosystem integration and pioneering agent-based payment and AI computing consumption…..11 hours ago13.3KLarge Model Company Launches Smartphones to Compete with OpenAI: Step Stars to Unveil Its First AI Agent Terminal on July 13thJieyue Xingchen to hold July 13 conference themed ‘True Agent in the Agent Era,’ unveiling next-gen agent terminal products, possibly including AI terminal brand, agent system, and first AI agent phone. Aligns with OpenAI’s push for new AI terminals, signaling industry acceleration in agent hardware…..11 hours ago10.9KTencent to Take Over Manus and Become the Largest Shareholder, Valuation of $2 Billion to Focus on Intelligent Body EcosystemTencent is negotiating to become the largest shareholder of Singapore AI startup Manus, after Meta’s $2B acquisition was ordered revoked. Existing backers including Tencent, ZhenFund, and Sequoia China plan to support the revocation at the same valuation, with Tencent subscribing for the most shares and possibly retaining a minority stake…..12 hours ago13.1KGoogle Releases LiteRT.js: AI Inference Runs Up to 3 Times Faster with WebAssembly Hardware AccelerationGoogle released LiteRT.js, replacing TensorFlow.js’s JavaScript kernel. It uses WebAssembly and deeply integrates WebGPU and WebNN hardware acceleration, significantly improving AI/ML runtime efficiency in browsers. WebAssembly enables near-native high-performance computing, optimizing in-browser inference speed…..13 hours ago14.4KGPT-5.6 Officially Released Today: Three Pricing Tiers Outperform Competitors, AI Development Enters a New Era of Cost-EffectivenessOpenAI releases three models in the GPT-5.6 series: flagship Sol, balanced Terra, and cost-effective Luna. These models cover ChatGPT, Codex, and API within 24 hours. The lowest pricing is $1 per million input tokens and $6 per million output tokens, just one sixteenth of competitors’. Performance comprehensively surpasses Claude, focusing on programming and cybersecurity, with the core selling point being extreme cost-effectiveness.15 hours ago20.3KMusk publicly admits he was wrong: Anthropic is the leader in AI, I won’t pull its serversMusk rarely admits misjudging Anthropic, now hailing it as an AI leader and pledging uninterrupted compute supply—a stark reversal from last year’s dismissive stance. SpaceX has already leased compute to Anthropic, highlighting deepening interdependence…..15 hours ago12.2KMiniMax Completes New Round of Financing Worth 16 Billion Hong Kong Dollars; Founder Yan J