A knowledge-sharing platform for Sohu's technology products. As a leading Chinese internet brand with media, video, search, and gaming services and over 700 million users, Sohu continuously drives tech innovation and practice. We’ll share practical insights and tech news here.
The article describes how an AST‑based code‑context service, wrapped by an MCP middleware and built with Spring AI and Eclipse JDT, supplies structured Java code information to large models, addressing the context gaps of diff‑only AI code review and improving accuracy through concrete examples and evaluation.
When sharding databases, independent auto‑increment IDs can collide, causing data integrity issues; this article explains why the conflict occurs and walks through five practical solutions—including UUID, custom auto‑increment steps, segment mode, Snowflake, and Meituan Leaf—detailing their trade‑offs, performance, and implementation steps.
Harness Engineering, the set of systems that surround and extend a large‑language‑model‑based agent, determines real‑world performance far more than the model itself, and mastering its six‑layer architecture, bottlenecks, and practical rollout steps is essential for AI‑agent development and interview preparation.
The author recounts how a bare‑bones Claude implementation missed edge cases, then introduces ten open‑source Claude Code Skills that embed a disciplined workflow—brainstorming, planning, execution, TDD, systematic debugging, parallel reviews, git worktrees, and diagram generation—showing installation commands, real‑world examples, and concrete benefits.
WasmGC introduces native garbage‑collection support to WebAssembly, enabling Kotlin and Dart to run as true native languages on the web, dramatically shrinking bundle sizes, speeding startup, eliminating memory‑leak cycles, and improving interop, with full browser support now standard across Chrome, Firefox, and Safari.
This tutorial explains the three-stage evolution from Prompt Engineering to Context Engineering and finally Harness Engineering, detailing their motivations, core components, practical implementations, and why stable, end‑to‑end AI agents require a full harness to manage tasks, context, tools, execution, state, and error recovery.
This article explains the architecture and evolution of a feed push system, covering the basic push model, its early implementation, performance trade‑offs, and a series of optimizations—including lazy loading, length control, delayed push, active‑inactive fan filtering, and hot‑cold separation—to improve read efficiency, reduce storage costs, and handle massive fan bases.
WebAssembly extends JavaScript by handling heavy computation, binary processing, cross‑language reuse, and secure sandboxing, making it a practical engineering tool for performance‑critical, privacy‑sensitive, multi‑platform, and plugin‑based web applications while complementing rather than replacing JavaScript.
In early 2026 the AI community witnessed a sharp shift away from Model Context Protocol (MCP) toward CLI‑first toolchains, as engineers highlight token inflation, fragmented authentication, and loss of composability in MCP, while praising the low‑friction, text‑based, and easily debuggable nature of command‑line interfaces for building robust AI agents.
This article details the design and implementation of an AI‑powered automated code‑review system for GitLab Merge Requests, covering background problems, layered architecture, diff parsing, prompt engineering, comment management, rate‑limiting, concurrency control, and the measurable improvements achieved.
