#AI编程
9368 related articles
Doubao Work Hands-On Test: Turn Three Messy Tables into a Reporting Dashboard with Zero Code
A Bilibili creator tests Doubao Work: turn three messy tables into a reportable business dashboard with zero code, covering the six-step delegation method, prompt iteration tips, and three red lines before delivery, plus a complete learning path.
Breaking Into LLM Development From Zero: A Four-Stage Learning Roadmap
Breaking down a four-stage LLM systems course from Bilibili: job-role awareness, core Prompt/RAG/Agent skills, engineering delivery, and job-hunting—how a beginner should plan a path into LLM development.
The Complete WorkBuddy Guide: How a Desktop AI Agent Does Your Work for You
A beginner's guide to the WorkBuddy desktop AI agent: how it differs from traditional Q&A AI, a four-dimensional comparison with Codex on barrier to entry, ecosystem, cost, and extensibility, plus practical scenarios like file management, document processing, and scheduled reporting, and the credit mechanism.
Making AI Comic Dramas with Codex: A Zero-Basics Fully Automated Production Workflow
A step-by-step Bilibili tutorial breaks down how to build a fully automated AI comic drama workflow with Codex AI Agent — covering scriptwriting, prompt generation, character design, storyboarding, and AI video generation in five steps, accessible even to beginners.
Fei-Fei Li's 80-Page Agent AI Survey: A Complete Guide to AI Agent Learning
Fei-Fei Li and researchers from Stanford and Microsoft release an 80-page Agent AI survey mapping agent theory, capability modules, and application challenges.
WorkBuddy: A Step-by-Step Guide from LLMs to Agent Platforms
A beginner-friendly WorkBuddy guide explaining the four-layer structure of LLMs, Harness, agents, and FDE deployment, with hands-on scenarios and a learning path.
RAG and Agent Interview Essentials: A Full Walkthrough of Retrieval-Augmented Generation
A systematic breakdown of the complete RAG retrieval-augmented generation workflow: from offline knowledge base construction via document loading, chunking, and vector storage to the online stage of query rewriting, vector retrieval, reranking, and generation, covering high-frequency interview topics and core value analysis.
Tencent MiaoYing Agent Hands-On: One Canvas for the Entire AI Video Workflow
Bilibili creator tests Tencent MiaoYing Agent: enter an idea, and the Agent auto-generates story, character, and storyboard cards and batch-creates video, all edited and exported on one canvas.
Harness Engineering: The Secret to Tripling the Same Model's Score
Harness engineering is core to AI agent development: the same model's score rises from 13.3% to 38.3% due to surrounding code. This article breaks down the seven components of a harness—loop, actions, context, memory, permissions, verification, and observability—with a refund agent example for practical methods and improvement paths.
OnlyMCP: One MCP Server to Unify Your Memory, Skills, and Agents
OnlyMCP unifies skills, agents, and memory behind a single MCP server. Claude Code, Cursor, and ChatGPT clients connect once for cross-device use, with session visibility and permission revocation.
One MCP to Connect All Devices: Ending AI's Forgetfulness
AI always forgets context? MCP proposes a unified protocol to connect all tools and devices, letting agent knowledge persist as memory and load on demand. This article examines its core claims, architecture, and what remains to be verified.
Merge CEO Speaks Out: Executives Who Don't Understand AI Are the Organization's Ceiling
Merge CEO Shensi Ding shares her hands-on experience with Astra: executives are the ceiling of organizational AI fluency; Astra excels at sustained multi-task execution and common-sense completion, enabling non-engineers to build tools autonomously.
Merge CEO Shensi Ding: AI Fluency Starts at the Top
Merge CEO Shensi Ding shares insights on AI fluency starting at the top: how Astra helps non-engineering team members independently build workflows, boosting organizational efficiency and innovation speed.
AI Agent Control in Practice: A Guardrail Demo for MCP Servers
Using the Acme Demo Corp. sample workspace, this demo shows how to configure guardrails for AI Agents calling MCP Tools: role-based tool access, disabled tools invisible to the Agent, field-level email masking, request limits with graduated Warn-to-Block responses, and Agent identity and delegation management.
Generate App Store Screenshots Automatically with an AI Coding Assistant: An Open-Source Tool Hands-On
App Store Screenshots is a free, open-source AI coding assistant skill for Claude Code, Codex, and Cursor, with 18 styles, Style Lab comparison, and multilingual iPhone/iPad/Mac/Android export—all running locally.
OpenLLM: Let AI Tokens Flow Freely Across Tools
OpenLLM is a developer tool that lets AI tokens flow freely across tools, sharing subscriptions, API Keys, and local models between Codex, the ChatGPT app, and local models, with automatic failover when quotas run out. This article analyzes its core features, use cases, and market feedback.
licenseal: Block Incompatible Open Source Licenses Before Release
licenseal is a free open-source CLI supporting 10 ecosystems. It reads lockfiles and registry metadata to check license compatibility of all dependencies, including transitive ones, without installing packages, giving CI a pass or block verdict.
GPT-6.1 Terra Hands-On: One-Prompt Websites, and the Trade-Off Between Speed and Cost
A hands-on YouTube creator test of GPT-6.1 Terra: one-prompt website building for North Star Coffee, a 57.6% Agents Last Exam score, comparisons with Sol and Claude, and real quota costs.
MCP vs. o3 Agents SDK: Why Gemini + Antigravity Wins Across the Board
Anthropic MCP is called AI's USB-C, while OpenAI's o3 Agents SDK, though strong at reasoning, is stuck in a cloud sandbox. Builder Wayne Stevenson explains why production favors Gemini + Antigravity multi-agent architecture for universal tool compatibility and local collaboration.
The Essence of AI Agents: One Loop, Five Core Steps
What is the essence of an AI Agent? This article explains its core loop: send the conversation to the model, check for tool requests, execute tools, and update the conversation until an answer is reached. Learn the three components and max turn limits.