The world of programming is changing fast as AI coding agents become a regular part of how software gets built. This weekly update shows that we're moving from early experiments to real, everyday use of AI helpers in coding work.

A major survey found that over 90% of engineering leaders plan to expand their use of AI coding tools in the near future. This isn't just a small trend - it's becoming the new normal for programming teams around the world. The numbers show we're at a turning point where AI tools are moving from "nice to have" to "must have" for most coding projects.

The growth predictions are stunning. Gartner projects that by 2028, perhaps 75-90% of software engineers will be using AI code assistants as part of their regular work. Just a year or two ago, fewer than 15% of programmers used these tools. This massive jump shows how quickly the industry is adopting AI helpers.

But there's an important point that experts keep making: these AI agents will work as assistants, not autonomous overlords. Think of it like a pilot using autopilot - the plane can fly itself in normal situations, but the pilot handles complex decisions and takes over when needed. Programmers will use AI to handle the straightforward 80% of coding work while they focus on the hard 20% that requires human creativity and judgment.

The types of AI helpers are becoming more sophisticated too. Researchers now talk about agentic workflows - these are AI systems that can make decisions, take actions, and coordinate tasks with very little human input. Unlike simple automation that follows fixed rules, these dynamic and flexible systems can adapt to real-time data and unexpected situations.

One breakthrough came from Zhipu AI in China, which created a system called ComputerRL. This AI can actually control computers the way humans do - clicking buttons, filling out forms, and navigating programs. On a difficult test called the OSWorld benchmark, it achieved a 48.1% success rate, which beat advanced models from OpenAI (42.9%) and Claude (30.7%). This shows AI is getting much better at using regular computer programs instead of just writing code.

The business world is taking notice of these changes too. Companies are realizing they need new policies and rules for AI agents that can make thousands of decisions per day. Unlike human programmers who might make dozens of choices daily, agentic systems operate at machine speed and scale. This means businesses need guardrails, ground truth, and governance encoded as code to make sure AI agents don't make dangerous mistakes.

Looking ahead, some experts think we're heading toward the end of platform dominance. Right now, most people use big platforms like social media sites or cloud services because building custom software was expensive and slow. But AI makes time very, very, very cheap when it comes to creating software. This means custom applications will become the norm, not the exception.

This shift could be huge for how we use technology. Instead of everyone using the same few big platforms, people and companies might have AI create exactly the tools they need for specific tasks. Why use a generic tool when a perfect custom solution costs almost nothing to make?

The improvements coming soon will make AI coding helpers even better. Companies are working on better reliability (fewer mistakes), better alignment with team conventions (following your company's coding style), and more seamless integration into dev environments. GitHub is already rolling out Copilot X features that integrate chat and task automation right in the code editor.

All these changes point to a future where programming becomes more creative and strategic work, with AI handling the routine tasks that take up so much time today.