Tencent AI Lab
AI overview
Sign in and the AI will write an overview from our coverage.
Headlines · 3
- Tencent reportedly pivots toward AI-native agents across its ecosystem
Tencent may be making a fourth AI strategy shift, moving away from betting on a standalone consumer AI app toward embedding agents across WeChat, mini programs, enterprise services, and other existing products. 木禾商业财经, republished by TMTPost, cites Ma Huateng’s “陆战变海战,陆军变海军” remarks at Tencent’s 28th-anniversary LinkTime meeting as describing a division between traditional businesses and AI-native work. The report says WorkBuddy and WeChat’s Xiaowei would lead the latter, while Hunyuan continues to develop agent, multimodal, and tool-calling capabilities.
钛媒体 · 🔥 6 - Deheng Ye joins ByteDance Seed as foundation model researcher
Deheng Ye updated his personal homepage to confirm that he is a researcher at ByteDance Seed working on foundation models. His work history lists 2017–2026 at Tencent, where he was R4 partner and technical lead for Honor of Kings JueWu, followed by ByteDance in 2026. The homepage also lists a SeedXxx model marked “P&C for now,” alongside the ui-voyager, omni-reasoner and proact generative-media agent systems, but their details and release status remain unclear.
InfoQ 中文 AI&大模型 · 🔥 0 - Cai Deng: post-deployment self-evolution is AI's fourth scaling axis
Cai Deng, who is joining Shanghai Jiao Tong University's AI institute, told a DeepTech interview that RAG-style external memory and ever-longer context windows both have ceilings, and that genuine continual learning ultimately requires models to update their own parameters while in use. He calls the stage after pretraining, post-training and inference-time scaling "Post-Deployment Scaling," arguing its output is not a stronger general model but one tailored to a specific person or organization. He also calls in-context learning the "endgame" learning method, with scalability as the open problem: internalize long contexts into weights in segments, then clear the window so learning can continue indefinitely.
MIT科技评论中文 · 🔥 0
Experience and discussion from the community
Share my Tencent AI Lab experienceAsk about Tencent AI Lab
Nobody has shared their experience with Tencent AI Lab yet.