An experienced software developer, drawing on two decades in the industry and a background with companies positive about Artificial Intelligence tooling, offers a detailed evaluation of phoenix.new as an AI-powered coding assistant for Elixir and Phoenix LiveView projects. The tool is praised for its deep understanding of Phoenix-specific concepts such as LiveView and Presence, outperforming most other automatic coding solutions in the domain. Users can rapidly generate working prototypes, with phoenix.new providing a contextual project plan and delivering a fully functional Phoenix app from just the first prompt—making it exceptionally efficient for initial project scaffolding.
The review, however, identifies significant friction in day-to-day development workflows. Key among these is the tendency for phoenix.new to overwrite files wholesale rather than perform targeted edits. This behavior results in manually introduced changes being regularly erased, undermining collaborative iteration between the developer and the coding agent. Additionally, the validation system is called out for frequent failures; component tests sometimes hang indefinitely, and the system often defaults to command-line validations instead of leveraging Elixir scripts or auto-generating proper test cases within the project’s test directory. Unlike competitors such as Claude Code, which proactively write scripts and test cases to validate logic and UI, phoenix.new requires explicit prompts to do so—and is more likely to misunderstand the underlying issue, leading to wasted credits and development time.
Beyond issues of collaboration and testing, the reviewer also flags diminishing output quality as projects progress. While the initial setup and MVP functionality are robust, phoenix.new frequently fills applications with static data rather than dynamic database content, requiring developers to intervene repetitively. Over time, both the quantity and quality of generated code decline, and addressing these shortcomings becomes costly due to the platform’s credit-based pricing model. Compared to alternatives like Claude Pro—which offer higher quality, more generous output, and fewer workflow interruptions for similar costs—the reviewer finds phoenix.new’s current value proposition lacking. Despite initial enthusiasm, recurring frustrations and higher-than-expected costs lead to skepticism about its suitability for long-term or production use, though the reviewer remains hopeful for rapid improvements as the platform evolves.
