I Tried Every AI Coding Agent... Here's My 2026 Setup

Summary

In this video, the creator shares their updated AI coding workflow for 2026, comparing it to six months ago. They discuss the tools and AI models they use, such as Cursor, Claude Code, and Gemini 3 Pro, emphasizing how each serves a specific purpose in their day-to-day work. The creator highlights the importance of parallel agent workflows, avoiding 'oneshot' coding, and using no-code tools like Mind Studio to increase productivity and involve non-technical team members. The video offers insights into how AI is reshaping their development process, improving both coding efficiency and UI design.

Takeaways

• 😀 The AI coding workflow is constantly evolving, and staying up-to-date with new tools and features is essential for maximizing productivity.
• 😀 Cursor is used primarily for tab-to-complete features, and though new functionalities exist, the user prefers sticking to the basics for consistency and comfort.
• 😀 GPT 5.2 Codec is slow but effective for long-running tasks and large code refactors, while Composer 1 is faster for shorter tasks, maintaining a smoother workflow.
• 😀 Gemini 3 Pro is highly favored for its superior UI/UX output, especially for tasks like design, where it outperforms other models like Claude or GPT.
• 😀 Parallel agent workflows are beneficial for tackling multiple tasks simultaneously, using tools like Branchlet to manage isolated Git branches.
• 😀 AI tools like Claude Code and Cursor offer significant flexibility for coding tasks, but using them efficiently requires a balance of planning, execution, and reviewing results.
• 😀 The focus has shifted from quick one-shot tasks to iterative, step-by-step coding to maintain control over the quality and reduce the need for constant rework.
• 😀 UI/UX design, especially for non-designers, has improved with AI tools like Gemini 3 Pro, allowing for better design output even from traditionally trained engineers.
• 😀 Low-code/no-code tools, like Mind Studio, have been integrated into the workflow to streamline non-technical tasks and enhance overall velocity, especially in growing teams.
• 😀 Incorporating AI coding tools, like the PR creator in Claude Code, helps automate repetitive tasks like updating documentation, keeping everything aligned with the codebase.
• 😀 Leveraging AI and low-code solutions isn't just about faster coding but ensuring the right tools are used for different types of tasks, whether coding, designing, or automating workflows.

Q & A

What is the focus of the video?

The video focuses on the creator's AI coding workflow for 2026, sharing their tools, techniques, and updates to their process compared to six months ago.

What role does the creator play in the tech world?

The creator is a coding YouTuber and the founder of a startup called Yori, a social media marketing tool. They also have a background in working at big tech companies.

What is the creator's primary focus outside of YouTube?

The creator spends 80-90% of their time building their startup, Yori, which helps businesses find and create high-performing marketing content.

Which AI tools does the creator primarily use in their coding workflow?

The creator primarily uses Cursor for tab completion and debugging, and Claude Code (especially Opus 4.5) as their primary AI coding agent. They also use models like GPT 5.2 CodeEx, Gemini 3 Pro, and Composer 1 for specific tasks.

How does the creator handle large refactoring tasks?

For large refactor tasks that require significant context and time, the creator uses GPT 5.2 CodeEx, despite its slower performance, because it is better suited for long-running tasks.

Why does the creator use Composer 1 for some coding tasks?

The creator uses Composer 1 for faster response times and quicker output, which helps them stay in their coding flow, even though the model is not as powerful as others like Opus 4.5 or GPT 5.2 CodeEx.

How does the creator use Gemini 3 Pro in their workflow?

Gemini 3 Pro is used by the creator primarily for UI and UX design tasks, as it provides high-quality design output that aligns with their aesthetic preferences.

What is the creator's approach to parallel agent work?

The creator uses parallel agents to handle multiple coding tasks simultaneously, leveraging tools like Branchlet and GitHub work trees to isolate changes and avoid merge conflicts, particularly when using AI models like Opus 4.5 and GPT 5.2.

How does the creator improve their app's UI design?

The creator collects screenshots from their favorite app designs, feeds them into Gemini 3 Pro to create a universal design language, and stores this design system in a cloud.md file for consistent reference during UI work.

What role does no-code/low-code play in the creator's workflow?

The creator incorporates no-code/low-code tools, such as Mind Studio, into their workflow to speed up app development and allow non-technical team members to contribute, especially for tasks like automating data processes and managing workflows.