On the Jukebox Podcast Nathan Wrigley interviews Corey Maass about how practical AI tools are changing daily work with WordPress. Corey, who has built for the web since 1997 and used WordPress since around 2010, describes a pragmatic, experiment-driven approach: adopt tools that clearly improve results, automate repetitive tasks, and leave human judgment where it matters.
Current stack and habits
Corey runs Cursor as an IDE, uses Claude Code to generate and scaffold code, and employs GitHub Copilot for in-editor reviews. He works with a Terminal inside Cursor, pushes changes to Git, and opens pull requests for human review. He calls himself an early adopter but not bleeding edge — he embraces tools that demonstrably help his workflow.
How the AI experiment began
A few months ago, while visiting a developer friend in Thailand, Corey began serious experiments with Cursor and Claude Code. He rebuilt an older project, Timerdoro, as a learning playground to test prompting styles and discover how granular instructions needed to be. What he learned is simple but important: models respond best to constraints and structure. Over time, models required less micromanagement, but early on the way you interact with them drastically affected outcomes.
Testbeds: projects used to learn and iterate
Corey keeps a couple of safe projects for trying out new tooling.
– Timerdoro: a long-lived productivity timer app he has rewritten across stacks. It serves as a low-risk place to test AI-driven rewrites and patterns.
– MexicanTrain.online: a Dominoes game he built during COVID that grew to thousands of users. He and a collaborator are rebuilding it with more formal processes — pull requests, test suites, and Copilot reviews — migrating from quick hacks to process-oriented development.
From time-saver to creative partner
Initially Corey used AI to save time: draft emails, scaffold placeholder code, or perform quick rewrites. Hallucinations in code were common a year ago, but recent model improvements reduced those issues and greatly expanded capabilities. Claude Code now scaffolds authentication flows and database structures quickly, so Corey shifted to using AI not only to save time but as a creativity engine. He experiments with prompts that introduce randomness or unusual constraints, such as translating text into another language and back or asking for childlike phrasing, to spark novel directions.
AI beyond code
Corey uses AI in non-code contexts too. For music production he fed mixes into Gemini for feedback on frequency peaks and interactions between kick and bass, getting actionable technical suggestions he might have missed. He also uses image generation to make playful editorial art or to refresh low-resolution headshots for his site, recognizing the tradeoffs between convenience and perceived authenticity.
Authenticity versus utility
Corey draws a clear line between functional software and artistic work. For clients, the priority is that software works; whether AI wrote parts of it is secondary if the result is maintainable and reliable. For creative work and live performance, authenticity and human connection remain central. He describes AI as a flip-a-card creativity tool — a way to inject serendipity, overcome blocks, and find new riffs — while still valuing human authorship in contexts where it matters most.
Humanizing tools
Corey and colleagues sometimes assign pronouns to tools, treating Copilot as female and Claude as male in casual conversation. He accepts a degree of anthropomorphism as natural in day-to-day interaction, while acknowledging that as AI blurs the line between tool and collaborator, cultural and ethical conversations will deepen.
Client work, pricing, and responsibility
When faced with an underfunded, hard-to-maintain codebase, Corey offered to rebuild at roughly half the former hourly cost because AI accelerated development. He was transparent about using AI and accepted responsibility for the final quality. Roles are shifting: developers are increasingly acting as product designers, project managers, and code reviewers, while AI picks up much of the typing and scaffolding. Humans still steer decisions, define architecture, ensure maintainability, and verify correctness. Clear file structure and consistent patterns remain important because they ease future comprehension for both human and AI collaborators.
Practical hacks and emergent workflows
Corey shared several pragmatic tricks. Screenshots are a surprisingly effective hack: many models can read text in screenshots, so he screenshots Copilot comments and feeds them to Claude to generate fixes or rebuttals. This reduces manual transcription and speeds iteration. He imagines tighter integrations where one model writes, another reviews, and a third reconciles comments, with humans stepping in for final QA and judgment.
WordPress-specific limitations and opportunities
WordPress is still a very UI-driven platform, meaning many admin tasks require clicking through interfaces and dealing with varied plugin screens. AI cannot yet reliably automate those UI interactions across the ecosystem. Plugins and patterns that expose machine-friendly formats, like ACF JSON exports, make AI work far easier. Corey predicts that stacks and plugins designed with predictable, text-based exports and APIs will be favored because they enable better AI integration and more repeatable workflows.
Outlook and concerns
Corey expects continued acceleration in certain development tasks, especially for new projects built from clean slates. Retrofitting legacy WordPress sites will remain slower because of UI idiosyncrasies and plugin complexity. He worries about broader societal effects: authenticity, misinformation, image misuse, education and grading when AI can both write and assess, and how research and creativity will shift in a human-AI hybrid world. Still, he is optimistic: when used responsibly, AI can be a coworker that extends human capability, enabling faster iteration and exploration of ideas.
Concrete examples
Corey pointed to practical outcomes: rebuilding Timerdoro and the MexicanTrain rewrite with AI scaffolding under human oversight; using Gemini to provide mix feedback and detect frequency issues; and employing image generation to produce editorial art and enhance headshots.
Advice for others
Be pragmatic. Try tools in low-risk projects first and adopt what clearly helps. Focus on outcomes and responsibility: clients care that software works, so be transparent and own quality. Expect roles to shift toward product thinking, testing, and review while losing some manual typing work. Anticipate platforms and plugins evolving to surface AI-friendly data shapes and APIs, which will make automation more reliable.
Where to find Corey
Twitter: @coreymaass
Company: gelform.com
Corey and Nathan close with a simple observation: AI is evolving fast, so workflows will continue to change. WordPress will likely evolve to offer better surfaces for AI to interact with sites, but human judgment, design sense, and responsibility will remain essential.