// edition · 2026-05-19
19 stories on AI dev tools, agents, and the coding stack — curated from the day's RSS haul by Agentic Dev's pipeline.
Anthropic's Claude Code plugin marketplace, as of May 2026, organizes extensions into four components: plugins (the distribution format), skills (Markdown-based prompt instructions), MCP servers (external data layer processes), and hooks (shell commands triggered by lifecycle events). The officia...
Claude Code, Anthropic's AI coding assistant, supports a layered extensibility system including CLAUDE.md project configuration files, custom slash commands, hooks, subagents, and MCP server integrations. A developer guide outlines how these features can be combined to automate code review, testi...
A developer guide published on Dev.to outlines Claude Code's extensibility features, including CLAUDE.md configuration files, custom slash commands, hooks, subagents, and MCP plugins. The guide describes how each layer functions and when to use it within a development workflow.
A tutorial describes building a stateful AI agent backend using FastAPI, LangGraph, and PostgreSQL to address production issues such as session memory loss and latency spikes under concurrent requests. LangGraph's persistent state graph replaces stateless API patterns by storing conversation stat...
Anthropic, IBM, and AI LABS independently presented talks arguing that hour-scale AI agent reliability depends on harness architecture, adversarial evaluator agents, and structured state handoffs rather than model improvements. Anthropic researchers Ash Prabaker and Andrew Wilson specifically pro...
GitHub launched remote control for Copilot agent sessions, now generally available on github.com and the GitHub Mobile app, with support also coming to VS Code and JetBrains IDEs. Developers can start a session locally using `/remote on` and then monitor, redirect, or approve actions from any dev...
Markagent is a Chrome extension that captures clicked UI elements and generates structured markdown prompts containing component names, file paths, CSS selectors, and screenshots for use with AI coding agents such as Claude Code, Cursor, and Codex.
A developer released a $99 bundle of Claude Code configuration assets, including 15 workflow skills, 5 automation hooks, 7 slash commands, and a CLAUDE.md template, sold as a single-developer lifetime license with a 7-day refund policy.
Guayoyo Tech rebuilt its website visual identity in under 48 hours using Claude Code as the design agent, modifying 62 files across 30 commits. The redesign replaced a generic cyan/Inter aesthetic with an amber palette, three-font editorial typography, asymmetric layouts, and dark/light mode sema...
A developer published an 8-level taxonomy for AI agent instruction systems, ranging from basic system prompts (L0) to self-improving agents (L7), submitted to the Hermes Agent Challenge on Dev.to. The framework applies across tools including CLAUDE.md, AGENTS.md, and .cursorrules, categorizing le...
Claude Pro and ChatGPT Plus both cost $20/month but differ in features: Claude Pro includes the Claude Code CLI and scores 64.3% on SWE-Bench versus ChatGPT Plus's 58.6%, while ChatGPT Plus offers image generation, voice mode, video via Sora, and Deep Research. Anthropic plans to separate program...
A developer overview identifies Cursor, Claude Code, and GitHub Copilot as useful for code navigation, automated refactoring, test generation, and documentation, while noting AI tools remain unreliable for complex debugging and domain-specific tasks.
DigitalOcean engineers achieved roughly 2x LLM inference throughput on Kubernetes by combining Managed NFS for shared model weights, jumbo frames with TCP buffer tuning, and a node taint to prevent a race condition between the network tuner and vLLM pods. The reference architecture, including Ter...
Guayoyo Tech redesigned its website in under 48 hours using Claude Code as the designer, modifying 62 files across 30 commits. The process involved providing the AI agent explicit design constraints via configuration files to replace a generic aesthetic with an amber palette, three-font editorial...
Simon Willison presented a lightning talk at PyCon US 2026 summarizing LLM developments since November 2025, noting the leading model changed hands five times among Anthropic, OpenAI, and Google, cycling through Claude Sonnet 4.5, GPT-5.1, Gemini 3, GPT-5.1 Codex Max, and Claude Opus.
Anthropic acquired Stainless, a New York-based startup founded in 2022 that automates the creation and maintenance of software development kits (SDKs). Stainless had previously provided its tools to companies including OpenAI, Google, and Cloudflare.
OpenAI and Dell announced a partnership to deploy Codex, OpenAI's AI coding agent, in hybrid and on-premise enterprise environments. The arrangement is aimed at organizations that require data to remain within their own infrastructure.
Business Insider reported that Google is internally testing an AI agent called Remy within a staff-only version of Gemini, describing it as a personal agent capable of taking actions on users' behalf across work, school, and daily tasks. Google declined to comment on Remy's existence.
Common AI integration failures at startups include poor data quality (cited in 85% of AI project failures), misallocated budgets toward customer-facing tools over back-office automation, and insufficient guardrails on autonomous agents. An estimated 42% of companies have abandoned most AI initiat...