In this video, the creator evaluates 17 AI coding assistants as of July 2025, with Claude Code emerging as the top performer, followed by Open Code and GitHub Copilot, while highlighting strengths, weaknesses, and cost considerations of various tools. The detailed, data-driven review emphasizes the importance of instruction-following and code quality, encourages community feedback, and helps developers make informed choices about AI coding assistants.
In this video, the creator shares their comprehensive evaluation of 17 AI coding assistants as of July 2025, highlighting new additions such as Codex CLI, Gemini CLI, Kilo Code, Open Code, IDER, and Warp.dev. The evaluation focuses on how well these tools follow detailed instructions and the quality of the code they produce, using a rigorous testing system involving unit tests and an LLM-based judging mechanism. The creator has transitioned to using Claude 4 for most tests due to its consistent performance and has adopted a points-based ranking system to better compare the assistants over time.
Claude Code emerged as the top performer, significantly outperforming others and achieving the highest scores in the evaluations. Open Code took second place, impressing with its execution quality and cost-effectiveness, while GitHub Copilot showed a remarkable turnaround, moving from near unusable to a solid third place, especially when using the creator’s own API key. Other tools like Trey, Zed, and Warp.dev also performed well, with Warp.dev gaining favor for its AI autocomplete features integrated into the terminal.
Some AI assistants, such as Gemini CLI and Codex CLI with GPT-4.1, scored poorly, with Gemini CLI’s performance being particularly disappointing despite the creator’s optimism for its future improvements. Kilo Code, a newer open-source tool that combines features from several other assistants, showed consistent but slightly lower scores compared to similar tools, raising some concerns about its marketing approach and actual innovation. The creator remains cautious about recommending Kilo Code but plans to monitor its development.
Augment Code, despite being a powerful AI coding assistant known for its context engine and task management, did not perform well in large-scale, specification-driven coding tasks. The creator uses it more for ideation and smaller coding tasks rather than big project implementations. IDER surprised the creator with better-than-expected results, while Void and some other assistants lagged behind in performance. The creator also noted the high cost of running some tests, especially with tools like Zed when using personal API keys.
In conclusion, the creator’s subjective ranking for July places Claude Code at number one, followed by Open Code and Augment Code, with GitHub Copilot making a significant leap in usability and quality. They emphasize that these rankings are personal and encourage viewers to form their own opinions. The creator also highlights the value of community engagement through their Discord server and invites feedback and discussion. Overall, the video provides a detailed, data-driven insight into the current landscape of AI coding assistants, helping developers choose the best tools for their needs.