This weekly update covers major developments in AI coding agents and the companies that make them. These smart helpers are changing how people write computer programs.

Google's Jules Goes Public

The biggest news was Google making Jules available to everyone. Jules is an AI agent that helps write computer code. It was in testing for many months, and now anyone can use it. During the testing period, thousands of programmers used Jules for tens of thousands of different coding tasks. Together, they created over 140,000 code improvements that were shared with other developers.

Jules got several important upgrades. It now uses Gemini 2.5 Pro, which is Google's most advanced AI brain. This makes Jules much better at planning how to write code before it starts working. Jules can also now work with both pictures and text at the same time. Google fixed hundreds of small problems that users found during testing.

Google now offers Jules in three different levels. The basic level is free for anyone to try. The Google AI Pro level gives users five times more usage, which is good for daily coding work. The Google AI Ultra level gives twenty times more usage and is meant for heavy users who need lots of AI help with their coding projects.

OpenAI Launches GPT-5 for Coding

OpenAI released GPT-5 this week, calling it their smartest AI model yet. This new model is much better at helping with computer programming than the older GPT-4. The company says GPT-5 has expert-level intelligence and can help with many different types of coding tasks.

GPT-5 is especially good at creating websites and apps. It can take one simple request and build a complete, working program. For example, if someone asks it to make a jumping ball game, GPT-5 can create the entire game with colorful graphics, sound effects, and smooth animations. The AI understands design principles like proper spacing, good typography, and attractive colors.

The new model also excels at debugging larger code repositories. This means it can find and fix problems in big, complex programs that have thousands of lines of code. Early testers noted that GPT-5 makes much better design choices than previous versions.

Business Challenges for AI Coding Companies

Despite the excitement around AI coding helpers, many companies in this field are struggling financially. A detailed report revealed that popular AI coding companies often lose money, even when they have many users.

The main problem is the high cost of using advanced AI models. Companies like Windsurf, which was valued at nearly $3 billion, actually spend more money providing their service than they earn from customers. This creates very negative gross margins, meaning they lose money on every customer.

AI coding companies face extra pressure because they must always use the newest and most expensive AI models. Programmers expect the best possible help, so companies can't use cheaper, older models. The competition is fierce, with popular tools like GitHub Copilot and Cursor fighting for users.

Some companies tried to solve this by selling themselves to bigger tech companies. Windsurf attempted to sell to OpenAI for $3 billion, but the deal fell apart. Later, the founders left to join Google in a deal worth $2.4 billion, while the remaining business sold to another company called Cognition.

Market Impact and Future Outlook

The week's developments show both the promise and challenges of AI coding agents. While tools like Jules and GPT-5 demonstrate impressive capabilities, the business model remains difficult.

The most successful path forward might involve companies building their own AI models instead of paying other companies like OpenAI or Anthropic. This could help reduce costs and improve profit margins. However, building good AI models requires significant investment and expertise.

For users, these developments mean access to more powerful coding helpers. Both Jules and GPT-5 can handle complex programming tasks that would take human programmers much longer to complete. The competition between companies is driving rapid improvements in these tools, benefiting programmers worldwide.