#AI编程
9355 related articles
Training a PPO Agent in UE5: 72,000 Nights of Guarding the Campfire
A developer shares a UE5 project training a PPO reinforcement learning agent to guard a campfire, covering ~72,000 nights of training and key issues in environment design, reward functions, and training efficiency.
HackerBooks: Using AI to Surface 16,500 Books Repeatedly Mentioned on Hacker News
HackerBooks catalogs ~16,500 books mentioned by at least two Hacker News commenters, each traceable to original comments. Built almost entirely by Claude, data downloadable.
AI Game Development in Practice: Letting Agents Fix Their Own Bugs, Clean Up Pixel Art, and Repair 3D Models
AI game dev in practice: Godot MCP lets an Agent fix bugs in three rounds; a 7-step script cleans 30K-color AI pixel art to a 9-color 32x32 sprite; headless Blender cuts 48K faces to 2,800 and fixes normals and origins. Includes MCP stage-gating and 3D cleanup budgets.
$0.53 Fine-Tuning for a High-Performance Model: A Complete Workflow Breakdown
A Reddit user fine-tuned a high-performance model for $0.53 and shared the workflow. We break down the cost structure, reusable process, and limits of low-cost fine-tuning.
Google Antigravity: Getting Started and Saving Money in Practice — Built-in Commands + Apprentice Asks the Master
A hands-on Google Antigravity guide: how built-in commands like BOST lift Gemini 3.8 Flash to near top-tier quality, the "apprentice asks the master" trick to drain web Gemini 4 quota, and the AgentLimb extension linking desktop and browser for low-cost AI development.
WorkBuddy for Beginners: Core Features and Agent Workflow in Practice
A beginner's guide to WorkBuddy: the three target user groups, conversational collaboration and workspaces, how experts, skills, and MCP connectors differ, and a learning path from AI comic drama practice to workflow transfer.
Xiaozhi AI Embedded Practical Project: Open-Source C++ ESP32 Project Becomes a New Learning Favorite
Bilibili creator details training camp's new project: ESP32 embedded practical tutorial based on Xiaozhi AI open-source project. Covers C++ engineering, ESP32 development, hardware abstraction, suitable for students, embedded beginners, and industrial software developers.
8,000 RMB Foreclosure Property Data Scraping Gig: A Hands-On Reverse Engineering of a JS Encrypted Parameter
A hands-on 8,000 RMB foreclosure property scraping gig: from packet capture to XHR breakpoint debugging, fully reconstructing the MD5 sign parameter logic, including concatenation pitfalls and code reproduction.
AI + SRC Vulnerability Hunting: A Step-by-Step Tutorial from Zero to Your First Bug
2026 Bilibili step-by-step AI+SRC vulnerability hunting tutorial covering fundamentals, vulnerability principles, AI tools, and hands-on hunting, guiding zero-based learners to their first bug.
AI Game Development in Practice: Three-Platform Build Publishing and Full Project Retrospective Workflow
A complete breakdown of AI game development lesson 13: Godot export notes for Windows, macOS, and Web, the itch.io publishing flow and AI copy constraints, plus how AI-assisted retrospectives produce a project constitution, prompt library, and pitfalls list.
Codex from Beginner to Master: A 6-Step Hands-On Workflow and Advanced Techniques
A six-step Codex tutorial based on a Bilibili hands-on course: project vs. thread isolation, routing tools for domestic models, three behavior modes, the three-step prompting method, plan and goal modes, agents.md expert configuration, and the insight that tools lower the bar while architecture sets the ceiling.
Building Mini Programs with WorkBuddy from Zero: Boundaries and Costs Explained
WorkBuddy from zero to WeChat Mini Program, Episode 1: capability boundaries, five suitable scenarios, four pitfalls to avoid, three cost accounts (time, account, cloud services), three entry points, and a copyable six-element prompt template.
FastAPI Beginner's Guide: Frontend-Backend Separation and Core API Design Concepts
Based on Bilibili's FastAPI beginner tutorial, this explains frontend-backend separation, FastAPI's Starlette foundation, API interfaces, and REST resource-oriented request method design to build core backend development understanding.
LangChain for Beginners: Understanding Model Integration and Tool Calling in One Lesson
LangChain beginner tutorial: analyzes the three major limitations of native large models, explains how LangChain connects models with external tools through chain-based thinking, covers Agent and RAG scenarios plus core operations such as message passing and tool calling, and includes learning advice.
Antigravity CLI: Free Access to Top Gemini and Claude Models in the Terminal
Antigravity CLI (agy) claims to offer free access to top models like Gemini 3.1 Flash and Claude Sonnet in the VS Code terminal, with installation in seconds. This article breaks down its workflow, the value of in-terminal model calls, and the security and sustainability questions behind the free claim.
Agentmash: An Open-Source Multiplayer AI That Stops Team Coding Agents from Colliding
Agentmash is an open-source multiplayer AI tool on Product Hunt that gives team coding agents a shared context layer, letting Claude Code, Codex, and others see each other's file changes, avoid conflicts, and @-mention teammates' agents like on Slack.
Termy: The AI Coding Agent Hidden in the MacBook Notch
Termy is a pull-down terminal tool that puts Claude Code and Codex coding agents into the MacBook notch, with real-time status, approvals, checkpoints, and a local server browser. This article analyzes its design and market signals.
Yardlet: An Open-Source Tool for Managing Codex and Claude Code on Mac
Yardlet is a free, open-source Mac app that unifies Codex, Claude Code, and other coding agents with a sidebar, approval cards, and a PR tab—without touching configs.
AwaitIO: Write a TypeScript Function, Instantly Get an Admin Panel
AwaitIO turns async TypeScript functions into internal admin panels. Each await io.input renders a form—zero frontend code, self-hosted, zero telemetry, free for 3 users.
Neurolume: Interactive Visualization of PyTorch/Keras Model Training
Neurolume is an open-source tool that interactively visualizes PyTorch/Keras model architectures in the browser and monitors training in real time with loss curves, overfitting checks, and weight-change animations.