Best of Claude — September 2025
- 1
- 2
Zed·38w
Claude Code: Now in Beta in Zed — Zed's Blog
Zed editor now offers Claude Code integration in public beta through their new Agent Client Protocol (ACP). This allows developers to use Claude Code as a native feature within Zed's high-performance editor, with capabilities like real-time code editing across multiple files, granular change review, and custom workflow definitions. The integration is built using ACP, an open standard that enables any AI agent to connect with compatible editors, and the Claude Code adapter is open-sourced for broader adoption.
- 3
Reinier·38w
Vibecode Sandbox
Vibecode Terminal offers a sandboxed environment for using AI coding agents without complex setup or infrastructure management. The platform integrates popular AI agents like Claude Code and Codex with GPT-5, providing features like run buttons, save functionality, and terminal access to simplify the development experience.
- 4
Collections·34wAnthropic Releases Claude Sonnet 4.5: State-of-the-Art AI for Coding
Claude Sonnet 4.5 achieves a 77.2% score on SWE-bench, positioning it as a leading AI coding model. Available on GitHub Copilot, VS Code, and JetBrains IDEs, it features enhanced memory management for tasks up to 30 hours, improved tool orchestration, and autonomous task handling. The model integrates with Snowflake Cortex AI and Amazon Bedrock for enterprise deployment, with pricing at $3 per million input tokens and $15 per million output tokens. Safety improvements include reduced sycophancy and better resistance to prompt injection attacks.
- 5
AI Coding Masterclass: From Beginner to Expert in 90 Minutes (Vibe Coding)
A video tutorial covering AI coding tools, specifically highlighting Claude Code and Codex as leading AI coding agents. The content appears to be structured as a comprehensive 90-minute masterclass designed to take viewers from beginner to expert level in AI-assisted coding.
- 6
Theo - t3․gg·36wIt's not just you (Claude did get dumber)
Anthropic's Claude models experienced significant quality degradation and reliability issues over several months, with multiple bugs affecting model intelligence going unnoticed and unacknowledged for extended periods. The company's lack of transparency, poor infrastructure reliability, and prioritization of research over user experience has created trust issues. Alternative providers like Google Vertex and AWS Bedrock offer better uptime for Claude models, and services like OpenRouter provide failover capabilities to avoid single-provider dependency.
- 7
GitHub Changelog·35w
Claude Opus 4.1 is now generally available in GitHub Copilot
Claude Opus 4.1 is now available in GitHub Copilot for Pro+ and Enterprise users across multiple IDEs including VS Code, Visual Studio, JetBrains, Xcode, and Eclipse. Enterprise administrators need to enable the new policy in Copilot settings, while Pro+ users can activate it through the model picker with a one-time confirmation prompt.
- 8
ThePrimeTime·35wClaude got dumber
Claude AI experienced performance degradation due to three infrastructure issues: context window routing errors that misrouted requests to servers configured for larger context windows, output corruption causing overemphasis on rarely produced tokens, and a top-k algorithm compilation bug involving mixed precision arithmetic. These technical problems compounded to create poor user experiences, but Anthropic clarified they never intentionally reduced model quality during peak times.
- 9
Hacker News·34w
Pairing with Claude Code to Rebuild My Startup’s Website
A non-technical founder shares their experience rebuilding their startup's website using Claude Code and AI coding agents. The process took weeks instead of months, allowing them to implement high-fidelity designs from Figma without learning to code from scratch. The workflow involved using VS Code, Claude Code CLI, GitHub CLI, and Figma's MCP server, following standard development practices like branching, pull requests, and code reviews. While powerful, the experience revealed several challenges including Claude's inconsistent response quality, file management issues with the Figma integration, and the need for constant human oversight to prevent the AI from making unrelated changes.
- 10
To AI or not to AI
A development team conducted a two-week experiment building an app using full AI assistance with Claude Code, but encountered significant challenges including lack of context awareness, poor code maintainability, workflow disruption, API hallucinations, and the 80/20 problem where AI handled basic tasks well but struggled with edge cases. Despite the frustrations, they continue using AI selectively for search, rubber ducking, code snippets, testing, and language tasks while maintaining control over the development process.
- 11
advanced-context-engineering-for-coding-agents/ace-fca.md at main · humanlayer/advanced-context-engineering-for-coding-agents
A detailed exploration of advanced context engineering techniques for AI coding agents, introducing "frequent intentional compaction" - a research/plan/implement workflow that manages context windows effectively. The author demonstrates how these techniques enabled shipping significant features to large codebases (300k LOC Rust projects) while maintaining code quality and team alignment. The approach emphasizes structured context management, high-leverage human review at key points, and treating specs as the new source code rather than just chatting with AI tools.
