Nathan Wrigley hosts Corey Maass on the Jukebox Podcast to discuss real-world AI tools and workflows for WordPress development. Corey, building for the web since 1997 and using WordPress since about 2010, describes how AI has reshaped his daily work: from coding to creativity to client projects.
Current stack and approach
– Corey uses Cursor as an IDE, Claude Code to generate code, and GitHub Copilot for code review. He opens Terminal inside Cursor and pushes changes to Git with pull requests for review.
– He’s pragmatic and an early adopter but not bleeding edge—he adopts tools that demonstrably help his workflow.
How his AI journey began
– In March, while visiting a developer friend in Thailand, Corey started experimenting seriously with Cursor and Claude Code, rebuilding an old product (Timerdoro) as a playground to learn patterns and how granular prompts should be.
– He discovered how much interaction style matters: give the model constraints and structure versus broad conceptual prompts. Over time models required less micromanagement.
Projects used as testbeds
– Timerdoro: a long-lived productivity timer app rewritten multiple times across stacks serves as a safe place to test AI-powered rewrites.
– MexicanTrain.online: a Domino game Corey built during COVID that grew to thousands of users. He’s rebuilding it with a friend, introducing pull requests, testing suites, and Copilot reviews—moving from “make a quick change” to more process-driven development.
From time-saver to creative collaborator
– Early AI use was mostly about saving time: drafting emails, placeholder code, or quick rewrites. Hallucinations were common a year or so ago, especially in code.
– Recent improvements reduced hallucinations and expanded capabilities: Claude Code can scaffold auth flows and databases quickly. Corey shifted from only time savings to using AI as a creativity engine—introducing randomness, idea generation, and alternative phrasing (e.g., “translate to German then back” or “write like a 5-year-old”) to find novel directions.
– He applies AI beyond code: for music production he used Gemini for mix feedback (frequency peaks, kick/bass interaction), which provided technical suggestions he otherwise might not have noticed.
Authenticity and “the point”
– Corey draws a distinction: for software clients, functionality matters most—who or what wrote the code is secondary; for art (music, live performance) authenticity and human connection remain central.
– He views AI as a “flip-a-card” creativity tool (like Brian Eno’s Oblique Strategies): a way to inject serendipity and overcome creative blocks.
– Examples: using image AI to create fun magazine art (editor’s head as Sasquatch) or to refresh his own site photo—effective and useful but raising questions about perceived authenticity.
Humanizing AI and relationships
– Corey and colleagues sometimes assign pronouns to tools (e.g., Copilot as “she,” Claude as “he”). He’s comfortable with casual anthropomorphism and finds it natural in day-to-day interaction.
– He notes cultural and ethical conversations will deepen as AI blurs lines between tools and collaborators.
Client work, pricing, and responsibility
– A client had underfunded, hard-to-maintain software. Corey proposed rebuilding at roughly half the previous hour-based cost because AI accelerated development. He disclosed AI use and accepted responsibility for final quality.
– Roles are shifting: developers become product designers, project managers, and code reviewers. AI handles more of the typing and scaffolding; humans steer decisions, design architecture, ensure maintainability, and verify correctness.
– Maintaining clear file structure and patterns still matters: while AI can place files, logical organization helps future comprehension for both humans and AI.
Practical hacks and workflows
– Screenshots are a surprisingly powerful hack: many models can read screenshot text, so Corey screenshots Copilot’s comments and feeds them to Claude for fixes or rebuttals. This short-circuits typing and streamlines iteration.
– He envisions tighter tool integrations where models coordinate: Claude writes, Copilot reviews, Claude reconciles comments—ideally with minimal human intervention until final QA.
WordPress-specific notes and limitations
– WordPress remains “clicky-clicky”: many tasks require UI interaction. AI can’t yet reliably perform UI clicks across varied plugins and admin screens.
– ACF’s JSON export is AI-friendly because it’s text-based; many other plugins aren’t. Future plugin design that intentionally exposes capabilities and data shapes (APIs/JSON) will enable better AI integration.
– Adoption will favor stacks that make AI interaction predictable. As plugins lean into AI-friendly formats, some stacks may become preferred.
Future outlook and concerns
– Corey expects more acceleration: certain development tasks will become much faster, especially new-project work from clean slates. Retrofitting older WP sites is slower due to UI and plugin idiosyncrasies.
– He worries and contemplates broader societal impacts: authenticity, misinformation, image generation misuse, educational integrity (AI writing/AI grading), and how scientific or creative breakthroughs might emerge from hybrid human-AI thinking.
– Optimistically, AI becomes a coworker that extends human capability—enabling people to produce more, iterate faster, and explore ideas they might not have reached alone.
Practical examples of AI in action
– Revamped Timerdoro and MexicanTrain rebuilds with AI scaffolding and human oversight.
– Music mix feedback from Gemini that provided technical, actionable guidance about frequency peaks and kick/bass ducker use.
– Image generation to craft fun editorial art and to enhance personal site headshots when originals were low res.
Advice and closing thoughts
– Be pragmatic: adopt tools that clearly help you, and experiment in low-risk projects first.
– Focus on outcomes and responsibility: clients care that software works; be transparent and own quality.
– Expect roles to shift: more product thinking, testing, and review; less manual typing for many developers.
– Anticipate plugin and platform evolution to surface capabilities that AI can reliably use.
Where to find Corey
– Twitter: @coreymaass
– Company: gelform.com
Corey and Nathan close by noting AI’s rapid evolution means workflows will continue to change, with WordPress eventually offering better ways for AI to interact with sites while human judgment remains crucial.