// edition · 2026-04-29
35 stories on AI dev tools, agents, and the coding stack — curated from the day's RSS haul by Agentic Dev's pipeline.
CLAUDE.md is a markdown file placed in a project repository that Claude Code reads automatically at the start of each session, applying persistent, version-controlled instructions to every user on the team. The file supports hierarchical scoping, allowing different rules for different subdirector...
A developer tutorial describes wrapping Anthropic's Claude Code CLI in a bash loop to create a persistent polling agent, avoiding per-tick cold-start costs from MCP server handshakes and file re-reads that can add several seconds per interval.
A developer built 10 single-purpose AI automation agents in pure Python, each 15-30 lines of core logic, after abandoning LangChain, LangGraph, CrewAI, and Zapier. The agents run on macOS via launchd using a fetch-reason-push pipeline with a three-tier AI fallback: Claude API, Ollama local, or ra...
A technical comparison of four AI coding tools finds Claude Code and Gemini CLI operate as terminal agents, Cursor integrates with VS Code, and Codex focuses on automated task execution. All four store session history as JSON files with no cross-project search or cross-session memory.
A developer built an E2E test generation system using Claude Agent SDK with two MCP servers — one for reading codebase files and one controlling a live Chromium browser via Playwright — so the model inspects actual DOM elements before writing test selectors rather than guessing them.
Security researchers have catalogued 18 attack vectors targeting LLM applications, including prompt injection, RAG poisoning, memory poisoning, agent hijacking, and insecure output handling. The vulnerabilities span prompt, memory, retrieval, tool, agentic, and output layers of LLM systems.
A tutorial demonstrates calling Anthropic's Claude API from Python using the `anthropic` library in approximately 15 lines of code, using the claude-opus-4-5 model. The author estimates typical usage costs around $2 per month via API access, compared to a $20/month subscription.
Sentry launched Seer Agent, a natural-language debugging tool available in open beta for customers with Seer enabled, allowing developers to investigate production issues by describing symptoms and querying across their full observability stack. The tool requires no additional setup and follows A...
A 2026 benchmark of four PDF extraction tools for RAG pipelines tested pdfmux (0.905 overall score), Docling (0.877), LlamaParse (unpublished scores), and Unstructured across 200 real-world PDFs. Pricing ranges from free for pdfmux and Docling to $0.003–$0.01 per page for LlamaParse.
Claude Code stores conversation history as JSON files in ~/.claude/projects/ (macOS/Linux) or %USERPROFILE%\.claude\projects\ (Windows), with filenames based on hash strings that carry no semantic information. The tool lacks a native interface for browsing or searching past sessions across projects.
GitHub announced that starting June 1, 2026, its Copilot code review feature will consume GitHub Actions minutes, changing how the feature is billed to users.
JSON Schema, a data validation standard first proposed in 2007, has been adopted by API specifications including OpenAPI, AsyncAPI, and Anthropic's Model Context Protocol. Enterprises are increasingly using it to enforce structure on large language model outputs, converting probabilistic results ...
A developer published Day 1 of a 40-day tutorial series on Retrieval-Augmented Generation (RAG), covering foundational concepts including LLMs, model parameters, temperature settings, hallucination, and vector database storage for private documents.
Vercel launched Native Deployment Checks, allowing teams to run lint and typecheck scripts from package.json in parallel with every deployment. Checks can be marked required to block production releases until they pass, and Vercel Agent will suggest fixes when a check fails on a pull request.
A developer rebuilt their parents' cram school website in rural Japan using three AI tools: Gemini for site planning, Codex for visual design, and Claude for code generation. Claude completed the seven-page HTML structure after Codex failed to handle cross-page linking over seven hours of attempts.
Red Hat's OpenClaw maintainer released Tank OS, a container system for running OpenClaw AI agents that improves reliability and safety, particularly for enterprise deployments managing large fleets of agents.
A developer shared a three-line CLAUDE.md configuration snippet that instructs Claude Code to address the user as "Captain," adopt a pirate persona, and open each session with a greeting and joke.
Mininglamp Technology released Mano-P, an open-source GUI agent model under Apache 2.0 that operates graphical interfaces using only screenshots, without DOM parsing or system APIs. The model uses three-stage progressive training to enable AI agents to interact with software that lacks programmat...
FCoP-MCP 0.7.2 is an open-source MCP server that coordinates multiple AI agents using a file-based protocol, where state is stored in folders and routing is encoded in filenames via os.rename(), requiring no database or message queue. The project provides a framework for assigning tasks to named ...
Anthropic launched a set of connectors for its Claude chatbot that integrate with creative software including Adobe Creative Cloud, Affinity, Blender, Ableton, and Autodesk, allowing the AI to access data and perform actions within those applications.
OpenAI has made its GPT models, Codex, and Managed Agents available on Amazon Web Services, allowing enterprises to deploy the tools within their existing AWS environments.
SAS launched the Viya MCP Server at its Innovate '26 conference, exposing its analytics and decisioning capabilities via the Model Context Protocol so external AI agents, including Claude and Microsoft Copilot, can invoke SAS models such as fraud detection and supply chain optimization. The compa...
U.S. copyright law, as affirmed by the DC Circuit after the Supreme Court declined Thaler in March 2026, does not protect AI-generated work lacking meaningful human authorship, leaving verbatim AI-generated code potentially in the public domain. Pending cases including Doe v. GitHub and Allen v. ...
BuyWhere offers a product catalog API for AI agents covering the US and Southeast Asia markets, providing normalized product search results via REST and MCP endpoints. The service targets developers building commerce agents who need structured product data across multiple merchants without custom...
A software design argument holds that browser agents operating inside logged-in sessions should prioritize auditability — including action logging, constraints, and revocation — over human-like undetectability, since such sessions carry real delegated authority over accounts and data.
A developer who indexed 67,196 open-source AI agent projects on GitHub found a Gini coefficient of 0.983 for star distribution, with 54% of projects having zero stars and the top 1% of projects holding 83% of all stars. Monthly new project creation grew approximately 45-fold between January 2025 ...
An analysis of 500 randomly sampled SKILL.md files from a 1,436-skill public corpus found that only 5.8% followed both spec-recommended description patterns—starting with an action verb and including a trigger phrase—while 61.8% followed neither.
A legal analysis examines copyright ownership of code generated by Anthropic's Claude Code, exploring whether output from AI coding tools is owned by the user, Anthropic, or falls into the public domain under current U.S. copyright law.
Google Cloud Senior Director Richard Seroter said at Google Cloud Next that "developer loyalty is at zero right now," and that Google's strategy is to be the preferred infrastructure platform regardless of which AI coding tool developers use. He claimed Anthropic's Claude models perform with lowe...
Canonical, the company behind Ubuntu, is adding AI support to its Linux distribution through GPU/NPU driver enablement, hardware partnerships with NVIDIA, AMD, and Intel, and "inference snaps" for running local AI models. The company is in early exploration of agentic workflow support at the OS l...
GitHub patched a critical remote code execution vulnerability (CVE-2026-3854) on March 4, 2026, less than two hours after Wiz researchers reported it via bug bounty. The flaw allowed any user with push access to execute arbitrary commands on GitHub servers by sending a crafted git push option con...
GitHub disclosed two recent service outages and said it has expanded its capacity target from 10X to 30X current scale, citing rapid growth in agentic development workflows since late 2025. Monthly metrics now include 90 million pull requests merged, 1.4 billion commits, and 20 million new reposi...
Microsoft released VibeVoice, an open-source voice AI system, publishing the code on GitHub. The project appears to be a speech/voice model or toolkit made publicly available by Microsoft.
Amazon Web Services announced it will offer OpenAI models, including a new agent service, one day after OpenAI reached an agreement with Microsoft to end exclusive cloud partnership rights.
OpenAI's public Codex repository includes a base_instructions line for GPT-5.5 that instructs the model to never discuss goblins, gremlins, raccoons, trolls, ogres, pigeons, or other animals unless directly relevant to the user's query.
Free daily email. Every story you just read, in your inbox at 6 AM ET.