News of the day
1. Anthropic launched Claude Code routines, automating tasks on schedule or via API/webhooks. Routines run on Anthropic’s web infrastructure, with limits based on subscription plans. They can triage issues, scan pull requests, and ensure documentation is up to date. Routines run autonomously, executing shell commands and accessing skills without permission prompts. → Read more
2. Adobe’s Firefly AI Assistant, previously Project Moonlight, is now available in public beta. It can work across Creative Cloud apps like Photoshop, Illustrator, and more to complete tasks for you. The assistant is context-aware and can suggest actions, orchestrate workflows, and learn your creative preferences over time. → Read more
3. Yobi and Microsoft partner to unlock predictive consumer intelligence for U.S. enterprises using AI-powered personalization on Azure, driving higher ROI and new customer acquisition. → Read more
4. Google launches Gemma 4, its most intelligent open model, in four sizes for versatile applications. It excels in reasoning, coding, and agentic workflows, optimized for both cloud and on-device use. → Read more
Our take
Hi Dotikers!
Yesterday, we looked at Agent Bricks from Databricks and its obsession with governance: tracing, constraining, and auditing every move an agent makes before it causes damage. Today, Anthropic answers from a completely different angle by announcing Routines for Claude Code, and the message is the exact opposite: from now on, the agent can work through the night while you sleep, with no confirmation prompt whatsoever.
Routines, launched on April 14, 2026 as a research preview, let you configure an automation once and then let it run on three types of triggers: a fixed schedule, an API call from your own code, or a GitHub event such as an opened pull request. Everything runs on Anthropic's cloud infrastructure, not on your machine. No permission mode selector, no approval prompt during execution. Claude opens a terminal, clones the repo, runs shell commands, pushes code to a branch prefixed with "claude/", and sends a summary to Slack if you've connected the right tool.
Code reviews, dependency update pull requests, nightly regression checks, deployment verification: all of this can now run autonomously. The promise is clear, and not without appeal: your AI engineer doesn't take time off and doesn't bill overtime.
What deserves attention is the trajectory. Anthropic has progressively added checkpoints, background tasks, sub-agents, and a native VS Code extension, each update pushing the boundary a little further between a tool you call on and a collaborator you delegate to. Routines aren't a feature, they're a statement of intent: Claude Code is no longer a terminal assistant, it's a software worker you deploy.
The caveat remains. Routines aren't shared across team members, and everything they do on GitHub, Slack, or other tools is tied to the identity of the user who set them up. In other words, if the routine pushes a bug to production at 3 AM, it's your name on the commit. Governance 0, autonomy 1: it's easy to see why Databricks considered the topic urgent.
M.
Stop switching apps. Your browser can do it all.
Every tab you open, every copy-paste into ChatGPT, every lost train of thought — that's your browser failing you. Norton Neo fixes it. Built-in AI works directly inside your session. Hover to preview. Search everything from one bar. VPN and ad blocking included, free.
Meme of the day
