#agentic-coding

#agentic-coding

Just like with a human developer, if you give your coding agent crappy instructions, if you don't give it proper guidance, if you don't take the time to point it in the right direction, you'll get bad code. If you don't keep an eye on things and continually nudge things in the right direction, you'll end up in the wrong place. No surprise.

With the new desktop app, users can orchestrate multiple agents and execute tasks simultaneously. Plus, you can design custom subagent workflows and schedule tasks that can automatically run in the background. The new app also easily lets you integrate projects with Google AI Studio, Android, and Firebase.

Legacy software, often described as 'big balls of mud,' has accumulated over decades, becoming difficult to maintain and understand. These systems rely on extensive teams to function, despite their outdated technology.

My website was not what potential customers would be looking for. Although what the site did was useful, no one in the age of agentic coding was going to use it. Instead, they would want some way to have their coding agent, or their build process, or some other automated thing, use my system.

Leading the pack has been the rise in agentic coding tools. These tools, such as Gemini Jules, Claude Code, and OpenAI Codex, are capable of writing entire programs and products. I put both Codex and Claude Code to the test, creating four plug-in add-on security products for WordPress using Codex, as well as a full-featured iPhone app using Claude Code.

If there's one thing I want you to take away from this article, it's this: testing harness is the most important thing for vibe-coding. Not prompt engineering, not fancy plugins, just constraining your AI outside AI toolchain. I'm calling it harness because it's not only tests. It's tests, types, linters, and any other automated checks you can put in place. The more you rely on AI, the more harness you need.

Andrej Karpathy is one of AI's guiding figures. He was a founding member of OpenAI and later served as Tesla's director of AI. He also coined the term "vibe coding," the AI-assisted coding movement that has taken software engineering by storm and was named Collins Dictionary's word of the year. In his "random notes from Claude Coding" - which are over 1,000 words long - Karpathy wrote about the changes to his own coding style.

One of the most notable innovations is support for compaction, a technique that enables the model to seamlessly manage multiple context windows and perform tasks involving millions of tokens within a single session. This enables large-scale refactors, extensive debugging rounds, and work cycles lasting several hours. GPT-5.1-Codex-Max is available immediately within the Codex environment in the CLI, IDE extension, and cloud; API access will follow later.

On Tuesday, Google finally launched Gemini 3, which the company claims is the "best model in the world for multimodal understanding and our most powerful agentic and vibe coding model yet." The claim is supported by benchmark data, crowd-sourced Arena results, and more advanced use cases that the chatbot has not previously been able to tackle. Also: Let Gemini make your next slide presentation for you - here's how

Anysphere has introduced Cursor 2.0, an update to the AI coding assistant that features the tool's first coding model, called Composer, and an interface for working with many agents in parallel. Both Cursor 2.0 and Composer were introduced October 29 by the Cursor team at Anysphere. Cursor is a fork of Microsoft's popular Visual Studio Code editor, downloadable at cursor.com for Windows, MacOS, and Linux.

Having a well-integrated type system that makes sense and gives you optimization potential for memory layouts is generally a good idea.