Dev Digest — April 14, 2026
🔥 HOT RELEASES
Claude Code Routines (Research Preview) Anthropic just dropped routines for Claude Code — configure a prompt, repo, and connectors once, then run on a schedule, API call, or event trigger. Runs on Anthropic's infrastructure, no laptop needed. Each routine gets its own API endpoint. Basically serverless AI agents. 🔗 X: https://x.com/claudeai/status/2044095086460309790
OpenClaw 2026.4.14 Smarter GPT-5.4 routing + recovery, Chrome/CDP improvements, subagents no longer get stuck, Slack/Telegram/Discord fixes, performance boosts. 🔗 X: https://x.com/openclaw/status/2044042546976883063 💻 GitHub: https://github.com/openclaw/openclaw/releases/tag/v2026.4.14
Resend Launch Week 6 — Day 2 Resend shipping all week. Day 2 features a new email editor. Track everything at their launch page. 🔗 X: https://x.com/zenorocha/status/2044055527358181684 🔗 Launch: https://resend.com/launch-weeks/6
🧪 INTERESTING REPOS
second-brain — Karpathy's LLM Wiki Pattern Implementation of Karpathy's viral second brain concept. One command install via npx skills. Ingests articles/papers/notes into structured Obsidian wiki with auto cross-references. Works with Claude Code, Codex, Cursor, Gemini CLI. 42k stars, trending #1 on GitHub. 🔗 X: https://x.com/coreyganim/status/2043777491886846275 💻 GitHub: https://github.com/NicholasSpisak/second-brain
Chrome DevTools MCP Open-source MCP server giving AI agents direct access to Chrome DevTools — console errors, network traces, performance audits. One npx command. Works with Cursor, VS Code, Windsurf, Gemini CLI. 🔗 X: https://x.com/suni_code/status/2043756631843057956 💻 GitHub: https://github.com/ChromeDevTools/chrome-devtools-mcp
Livepaper — Reproducible Research Papers Annotates claims in academic papers with inline prompts so AI agents can reproduce results. Caught slightly-off numbers in the author's own papers. Interesting approach to trust in research. 🔗 X: https://x.com/fjzzq2002/status/2043873408681648570 💻 GitHub: https://github.com/fjzzq2002/livepaper
Claude Code Best Practices (69 Tips) Boris Cherny's (Claude Code creator) best practices compiled into a repo. 42k stars. Key takeaways: don't micromanage (paste bug, say "fix"), /compact when context >50%, keep CLAUDE.md under 200 lines. 🔗 X: https://x.com/axiaisacat/status/2044023594942968070
🎥 WORTH WATCHING
Deploying to Internet Computer with Claude Code One prompt → Claude shipped a full decentralized blog to ICP mainnet. Rust backend, Internet Identity, stable memory, frontend. Video shows the entire experience. 🔗 X: https://x.com/ChrisYost_/status/2043773552575263015
AI is No Longer What You Think — 60 Second Explainer Ania Kubów on how AI has quietly shifted from what most people assume it is. Quick watch. 🔗 X: https://x.com/ania_kubow/status/2044068359038550088
Android Studio + Gemini Tips Google's thread on using Gemini in Android Studio as a real workflow tool, not just a chatbot. 3 practical tips from Google Developer Experts. 🔗 X: https://x.com/AndroidStudio/status/2044083257529004046
💡 TECHNIQUES & IDEAS
The Multi-Agent Coding Stack @ivanfioravanti's daily setup: 2x OpenCode (GLM 5.1), 1x Kimi Code (K2.6), 1x Gemini CLI (3.1 Pro), 1x Claude (Opus 4.6), 1x Cursor, 1x Hermes Agent (MiniMax M2.7). Running multiple coding agents in parallel on different models is becoming the norm. 🔗 X: https://x.com/ivanfioravanti/status/2043970443103596940
$2 USB-C LED for AI Agent Status Chinese student built a physical LED indicator for Claude Code — blinks when agents work, dark when idle. Simple hardware + software bridge. The "LED knows before you do." 🔗 X: https://x.com/noisyb0y1/status/2043900212331041172
MCP Second Brain Everywhere Your second brain now accessible from Claude, Claude Code, ChatGPT, OpenCode, Cursor, Gemini CLI, OpenClaw, Hermes Agent — every MCP-compatible agent can tap into the same knowledge base. 🔗 X: https://x.com/JoshuaIPark/status/2043907883990167784
🔮 EMERGING TRENDS
Claude Code Routines vs OpenClaw Multiple people noting Anthropic now has all the pieces that OpenClaw pioneered — scheduled agents, event-driven automation, repo-connected AI. The space is converging fast. 🔗 X: https://x.com/Hesamation/status/2044100493442453592
Gemini 3.1 Pro: Dead Last on Reasoning BridgeBench puts Gemini 3.1 Pro behind Grok 4.20, GPT 5.4, Claude Opus 4.6, Qwen 3.6 Plus, MiniMax M2.7, and GLM 5.1 on grounded reasoning. Google has the compute but can't ship a competitive model. Users cancelling $250/month AI Ultra subs. 🔗 X: https://x.com/bridgebench/status/2044024683473822136
GLM-5.1 Quietly Impressive Free Chinese model outperforming Google's flagship. Multiple devs praising it. The open/cheap model tier keeps getting better. 🔗 X: https://x.com/Mayhem4Markets/status/2043834029200318564
RISC-V Gets Free GitHub Actions Runners Linux Foundation's RISE Project launched free managed GitHub Actions runners on physical RISC-V hardware. Removes the biggest barrier for open source RISC-V contributors. 🔗 X: https://x.com/linuxfoundation/status/2043776495785533706
Full digest: https://digest.99.cooking