Spotify's Developers Stop Writing Code

Spotify, a big music streaming company, shared exciting news this week. The company's best programmers have not written any lines of code since December because AI agents are now doing all their coding work. Instead of typing code all day, engineers at Spotify now use an AI system called "Honk" that lets them give instructions to Claude, an AI assistant, and the AI writes the code for them.

One example shows how powerful this is. An engineer at Spotify can be on the bus going to work and use their phone to tell Claude to fix a bug in the iOS app—that is the app people use on iPhones. Claude finishes the work and sends a new version back to the engineer's phone through Slack, a messaging app, so they can approve it and send it to production before they even arrive at the office. Spotify says this system has made coding and getting new features to users much faster and easier.

A New Company Helps Manage AI-Written Code

A man named Thomas Dohmke, who used to lead GitHub (a place where programmers store their code), started a brand new company called Entire. The company received a huge amount of money—60 million dollars—to help solve a new problem that companies are facing. The problem is that AI agents now write code so fast that humans cannot read and check all of it.

Entire created a tool called Checkpoints that keeps track of why the AI agent wrote each piece of code and what instructions it was following. This helps programmers understand what the AI did and make sure it is correct. The tool works with Claude Code and other AI coding assistants. Apple also updated its code editor, called Xcode, to include Claude Agent so programmers can use the AI directly inside the tool they use every day.

OpenClaw: The Fastest-Growing AI Agent

One of the biggest stories this week is about OpenClaw, an open-source AI agent that is growing faster than anything on GitHub has ever grown before. OpenClaw is a helpful robot that can manage your emails, read your files, use your web browser, and control your social media accounts all from one place. It was created by someone named Peter Steinberger, and both Meta and OpenAI—two of the biggest AI companies in the world—are trying to buy it.

OpenClaw is so popular because it can do so many things without a human telling it every single step. It can read emails, do web searches, run tasks on a schedule, remember things from past conversations, and connect to services like Twitter, LinkedIn, and Facebook. Already, people are creating versions of OpenClaw written in different programming languages, like Python and JavaScript, and alternatives with fun names like NanoBot, PikoClaw, and IronClaw.

A New Way to Program AI Agents

Programmers are learning a completely new way to work called skills-in-the-middle. Instead of writing hard computer code, programmers now write simple instruction files in a format called Markdown. Think of it like writing instructions for a recipe instead of building a kitchen—much simpler and easier for AI agents to understand. Because of this change, system architects (people who plan how computer systems work) are becoming "skills architects" who describe what they want to happen instead of writing complicated code.

Safety and New Tools

As AI agents become more powerful and common, companies are worried about keeping them safe and secure. A group called the Coalition for Secure AI made a paper that lists almost 40 different dangers that could happen with AI agents. Major companies like Google, IBM, Microsoft, and NVIDIA are working together to keep AI agents safe. Meta joined this group this week, showing that even the biggest companies think this is important.

Google also created a new MCP Server (a special connection tool) that lets AI agents read Google's instruction manuals for its products. This helps AI agents answer questions more correctly and faster.

The Biggest Change: Humans Think, AI Codes

The most important change happening right now is that programmers are shifting from writing code to thinking about what to create. An expert at a company called Axios explained that because AI agents can now write code better and faster than humans, programmers should focus on imagining cool new features instead of worrying about how to code them. Features that used to take months to build now take just days.

However, this change is also creating new challenges. Because everything is changing so fast, some programmers feel tired from always learning new things. People worry that companies might not have a clear plan for all these AI-generated features. Even though humans can now code faster than ever, the real challenge is keeping up with how fast everything is changing.