AI Internet Daily: DeepSeek and Kimi reduce caching costs, Douyin Beanbao hotel entrance begins to charge commissions, Pinduoduo and Minda enter real-time retail, Meta open source Muse Glimmer

Let’s look at the ledger first. The entrance is becoming more and more expensive, and the cost savings will increasingly lead to changing owners. Anthropic turns on the Claude Code automatic mode by default, and the app store also begins to loosen the boundaries of distribution and payment. Meituan and JD.com were forced to the fulfillment efficiency table by Pinduoduo’s “fastest delivery tomorrow”. What I saw this time was not a single-point function upgrade, but the platform’s integration of assistants, payment, distribution and logistics into the same workflow. One less step, and more room for negotiation. Let’s take a look at the major events in the AI ​​and Internet tracks that happened in the past two days! 📌 A quick overview of major AI Internet events today 1. DeepSeek and Kimi reduce cache costs, NVIDIA promotes cross-model reuse 2. Alibaba Cloud AIDC delivery cycle is shortened to 100 days, and construction costs are simultaneously reduced