Sharing Zhihu tech posts and exploring community technology innovations.
This article details a six‑month, B‑side product team experiment that evolved from a failed AI tool trial to a mature human‑AI collaboration model, achieving 60‑70% automation of development tasks, cutting weeks of work, and establishing a knowledge‑base that lets AI act as a true team member.
This article reveals how Zhihu’s growth engineering team dissected activity‑page rendering bottlenecks, applied front‑end performance techniques such as early resource loading, image compression, pre‑rendering, and metric‑driven optimization (FCP, CLS, INP), and ultimately boosted page reach and conversion rates dramatically.
This article presents a novel minute‑level future‑reward framework with dual‑delay incentives, activity‑based attribution, multi‑task delayed modeling, and sequential streaming training that dramatically improves user retention prediction accuracy and real‑time performance in large‑scale recommendation platforms.
The article summarizes Zhihu’s technical talk on the ZhiLight large‑model inference framework, detailing model execution mechanisms, GPU load analysis, multi‑GPU parallel strategies, open‑source engine comparisons, compute‑communication overlap, quantization techniques, benchmark results, and future directions for scalable LLM deployment.
This article summarizes the key technical insights from Zhihu Direct Answer, an AI-powered search product, covering its product overview, RAG framework, query understanding, retrieval strategies, chunking, reranking, generation techniques, evaluation methods, and engineering optimizations for cost and performance.
The article recaps Zhihu's Cloud Native Innovation Forum where experts from AutoMQ, OceanBase, and Flashcat shared practical solutions on streaming data ingestion, unified database architectures, and AI‑driven observability, highlighting real‑world deployments, performance optimizations, and cost‑saving strategies.
The article analyzes the storage, sharding, backup, and operational pain points of Zhihu's self‑operated MongoDB cluster, proposes cloud‑based architectural and procedural solutions, and details a step‑by‑step migration plan that ensures zero‑downtime, improved stability, and cost efficiency.
The article details Zhihu's technical salon on large‑model engineering, covering the RAG‑based Zhihu Direct Answer system, the open‑source ZhiLight inference framework, prompt engineering, agent research, and future plans for integrating AI into product and community workflows.
This article explains TiDB’s distributed relational database architecture, compares Kubernetes‑based and physical‑machine deployments, and describes Zhihu’s platform‑driven approach—including TiDB Operator, TiUP, self‑service tooling, monitoring, backup, security, and FinOps—to enable efficient database management for both developers and DBAs.
This article presents a method for massive multi-label text classification on Zhihu content by combining a semantic retrieval model with a proprietary large AI model, detailing the challenges of large label spaces, model architecture, loss optimization, and experimental results showing significant accuracy gains.
