#AI编程
9239 related articles
TypeScript + Stateful Actors: Building a Highly Reliable Group Messaging Tracker
Crew Cast is an open-source TypeScript example using Stateful Actors and durable workflows to track group MMS delivery status, handling async events, webhook retries, and replies—ideal for scheduling and notifications.
Meta Acquires AI Coding Startup Conductor, Multi-Agent Collaboration in the Spotlight
Meta acquires AI coding startup Conductor, which focuses on multi-agent teams working on project tasks in parallel. This article examines the coding tool strategy behind the acquisition, its impact on developers, and the industry's competitive direction.
Replit Agent Lands on Slack: Write Code and Build Projects Right in Your Channels
Replit Agent is now available in Slack, letting users call the AI coding assistant in any channel to handle projects, connect integrated tools, and get instant answers. This article analyzes what the integration means for team collaboration and the shift in AI coding entry points.
20-Year Software Veteran Builds Robot Dog with Claude: Single Leg Already Hops
A 20-year software engineer used Claude to help design his self-built quadruped robot dog LocksleyFox. A single leg already hops, with walking planned in one to two months.
Replit Launches Government Edition: Empowering Civil Servants to Build the Future with AI
Replit launches Replit for Government, aiming to let civil servants directly use AI to build tools, reduce duplicative work, and accelerate government digital delivery. This article analyzes its product positioning, the AI adoption anxiety revealed in screenshots, and key issues to be resolved for government procurement.
Replit Weekly: Self-Driving Companies and the New Paradigm of AI Programming
Replit this week released Cloud Agents and ViBench 2.0, predicting every Fortune 500 company will become a self-driving company by 2030. We unpack natural language programming, security scaling, and long-horizon build benchmarks.
Building Internal Tools with Replit Saves Finance Team 150 Hours a Year
Clarisse built a finance tool with Replit, saving her team 150 hours a year. This article analyzes how low-code platforms empower business users as citizen developers, reshaping collaboration and efficiency.
Tired of Tutorials? This Free Game Lets You Learn Claude Code by Playing
Tired of endless Claude Code tutorials? Reddit developer launches free game Claude Code Quest with three difficulty tiers of hands-on challenges covering Claude Code basics, Skills, Hooks, Subagents, and MCP—plus open-source repo and prerequisites.
Workflow SDK 5.3.1 Atomic start Enables Idempotent Launch
Workflow SDK 5.3.1 introduces atomic start(), admitting a run and claiming a hook in one step. Duplicate launches throw HookConflictError with conflictingRunId. Learn the new API, createHook binding, and token-as-idempotency-key best practices.
Gemini Trains on Your Conversations by Default: Unless You Turn Off Keep Activity
A Hacker News discussion points out that paying for Gemini doesn't stop your conversations from being used for model training—what actually matters is the Keep Activity toggle. This article breaks down Google's documentation on human review and three-year retention, and lists the checks developers should do before pasting code.
Ripple-65B Preview: U.S. Open-Weight Model Tops Agentic Coding
Ripple-65B Preview debuts on Hacker News, claiming U.S. open-weight SOTA in Agentic Coding. We analyze its 65B positioning, the competitive focus of agentic coding, and key points to watch in the preview stage.
Replit for Business: Building a Self-Driving Finance Team
Replit for Business proposes the "self-driving finance team" concept. This article analyzes how AI programming platforms empower finance teams to automate, lower technical barriers, and explores the trends and challenges of enterprise AI programming adoption.
ViBench 2.0 Released: A Long-Horizon Software Construction Benchmark, Top Models Fall Short of 65% Pass Rate
ViBench 2.0 launches as a long-horizon software construction benchmark. Agents build products from empty projects; top models GPT-6.1 Sol, GPT-6 Astra, and Opus 5.5 pass all checks only 49–65%, revealing the capability gap in long-horizon construction.
Cloud Agents Update: Anyone in the Company Can Contribute to the Codebase
Cloud Agents now lets anyone in the company contribute to the codebase, with engineer review and approval; it also supports agent messaging, new sessions, and a GPT-6.1 Terra model option.
Replit's Deep Slack Integration: Letting Code and Data Flow Through Conversation
Replit's deep Slack integration turns feature requests into Pull Requests in minutes, connects codebases, pulls BigQuery data, and creates Linear tasks—all inside conversations.
LangGraph Agent Cost Control: recursion_limit Counts Steps, Only Dollar Limits Manage Money
LangGraph's recursion_limit only limits steps, not costs. Learn how to implement dollar-per-minute limits at the gateway and solve two pitfalls: window reset and key sharing, with LangChain implementation details.
Datasette Agent 0.5a0 Released: Generate Database Apps Through Conversation, Edit in Real Time with Side Panel
Datasette Agent 0.5a0 is released with a new panel feature that displays the AI chat window alongside Datasette data pages. Combined with datasette-apps, it can generate and modify database apps in real time through natural language prompts.
LLM 0.37.1 Released: Community Collaboration Fixes Critical Bugs
LLM 0.37.1 is released, a bug-fix version completed by community contributors that improves the tool's stability. Learn the highlights and upgrade advice.
AI Directly Rewrites Model Weights: MIT Study Reveals New Black-Box Breakthrough
MIT research shows frontier AI models can directly read and rewrite other models' raw weights without gradient descent. Eight experiments reveal weight space programming and self-improving AI.
Building a Real-Time Voice Assistant Jarvis with Python: Full MCP Tool Integration
Follow developer Mike's tutorial to build a real-time voice assistant Jarvis with Python and Agora. Integrate Google Calendar, Sheets, Notion, and IMAP email via MCP, and customize the LLM and voice for a personalized AI assistant.