The songs I liked in March 2025.
![Mystery White Boy (Expanded Edition) [Live] — Jeff Buckley · 2 songs](https://i.scdn.co/image/ab67616d0000b27366a82f1edd6013c64d923b41)
59 songs
The songs I liked in February 2025.
96 songs
A curated collection of design portfolios and creative collective sites. Featuring beautiful design work and community-driven creative projects.
Starting a collection of inspiring design portfolios and creative collectives. These represent thoughtful approaches to design practice - from individual portfolios that showcase beautiful work to collectives that prioritize community and collaboration.
Looking for more examples of design work that balances craft, creativity, and community values. If you know of portfolios or collectives doing interesting work, let me know.
A curated collection of developer blogs and technical writing that I find inspiring, educational, and thought-provoking.
A collection of developer blogs I regularly read for inspiration, learning, and staying current with the industry. These writers consistently publish thoughtful technical content that goes beyond surface-level tutorials.
Replace these placeholder links with your actual Chrome bookmarks!
A comprehensive reflection on everything I've learned so far about using AI tools in web development, creative projects, and problem-solving. From major successes to common pitfalls, here's what I wish I'd known when I started.
Over the past [TIME PERIOD], I've been deeply immersed in learning and working with AI tools across various projects - from web development and portfolio creation to creative problem-solving and technical challenges. This post serves as both a reflection on that journey and a resource for others embarking on similar paths.
Why document this now? Because the landscape is changing so rapidly, and the lessons I'm learning feel too valuable not to capture while they're fresh. This is meant to be a living document that I'll continue to update as I learn more.
Project: AI-Powered Portfolio
What worked: Using AI to help architect the entire site structure, from component design to content strategy
Key insight: AI excels at helping you think through user experience flows and technical architecture decisions
Live example: Portfolio Link
Project: Snake Game Easter Egg
What worked: Collaborative iteration between human creativity and AI implementation
Key insight: AI can help you implement fun, creative features that you might not have the time to fully research and build from scratch
Live example: Try the Snake Game
Project: Epic Astro Migration
What worked: Using AI to help navigate complex framework migrations and modern web development patterns
Key insight: AI can serve as an excellent pair programming partner for learning new technologies
Live example: Current Portfolio
Project: Editorial Process Optimization
What worked: Streamlining content creation workflows and establishing better creative processes
Key insight: AI can help you think systematically about creative workflows, not just implement them
Project: Symbiotic Survival - Minecraft Mutualism Mod
What worked: Vibe coding a complex game mod in hours using domain-specific AI workflows
Key insight: Claude adapts its approach based on domain—game dev gets incremental, placeholder-driven prototyping that prioritizes testing mechanics over polish
Full story: Building a Minecraft Mutualism Mod Dev Journal
Project: He's Going the Distance - Minecraft Difficulty Scaling Mod
What worked: Building on knowledge from previous mod to create gameplay-focused mod even faster
Key insight: Playtesting with another person revealed the mod worked better as co-op, leading to real-time difficulty tuning—AI tools enable fast iteration on game feel
Full story: Making a Distance-Based Difficulty Mod Dev Journal
The Problem: One of the most frustrating experiences I've had was when Claude got confused about Wagtail's scheduling features between different versions of the CMS.
What happened:
Working on a Django/Wagtail project
AI provided guidance based on outdated version information
Led to implementation issues and debugging headaches
Key Lessons:
Always verify version-specific information - AI training data can lag behind current software versions
Cross-reference with official documentation - Use AI as a starting point, not the final authority
Document issues publicly - GitHub issues help both you and others facing similar problems
Challenge: Keeping AI-suggested dependencies up to date and compatible
Solution: Developed a workflow for validating AI suggestions against current package versions
Tool: Created custom scripts for dependency checking before implementation
The Problem: Claude consistently exhibits specific patterns when working with Django migrations that can create workflow issues.
What happens:
Creates migrations on a per-app basis rather than considering cross-app dependencies
Often generates migrations without meaningful names, relying on Django's auto-generated naming
Frequently leaves migrations out of git commits, requiring manual review and addition
Better at including migrations when explicitly asked to use a specified migration name
Real-world impact:
Discovered during CCA Educational Platform development
Led to inconsistent migration history and deployment issues
Required establishing new workflows for migration management
Key Lessons:
Always specify migration names explicitly - "Create a migration named 'add_user_preferences'" vs. generic requests
Review git status before committing - Check that migrations are included in commits
Consider cross-app dependencies - Ask AI to analyze migration dependencies across the entire project
Establish migration naming conventions - Create clear patterns for AI to follow
Solution workflow:
Always request: "Create a migration named [descriptive_name] for [specific_change]"
After changes: "Show me git status and ensure all migrations are included"
Before deployment: "Review migration dependencies and execution order"
The Problem: Claude frequently attempts to create new labels when creating GitHub issues, rather than using existing repository labels.
What happens:
⎿ Error: could not add label: 'content-management' not found
⏺ Let me try creating the issue with just the labels that exist in the repository:
Real-world impact:
Issue creation fails when AI tries to use non-existent labels
Workflow gets interrupted by label creation errors
Requires manual retry with corrected label specifications
Key Lesson: Always specify to use only existing repository labels to avoid label creation attempts in issue-focused sessions.
Solution workflow:
Request: "Create a GitHub issue using only labels that already exist in the repository"
Alternative: "First show me the existing labels, then create the issue with appropriate ones"
Context separation: Handle label management in dedicated sessions separate from issue creation
Challenge: AI sometimes suggests solutions that work but aren't optimized for performance
Solution: Learning to ask better questions about performance implications upfront
Example: Asking "What are the performance considerations of this approach?" along with implementation requests
Game-changing features:
Contextual understanding of entire codebases
Ability to work across multiple files simultaneously
Excellent at explaining complex technical concepts
Best practices I've developed:
Always provide full context about the project structure
Ask for explanations along with implementations
Use it for code reviews and architecture discussions
Workflow that works:
Describe the problem and expected behavior
Share relevant code snippets and error messages
Ask for multiple potential solutions with explanations
Test implementations incrementally
Where AI excels:
Creating structured outlines (like this blog post!)
Generating comprehensive documentation from code
Brainstorming creative angles for technical content
Surprise discovery: AI doesn't just help you build things faster - it helps you understand things faster.
Example: Learning Astro's island architecture through collaborative explanation and implementation
Insight: The best AI interactions are conversations, not just commands
Unexpected benefit: Having AI as a brainstorming partner has made me more willing to attempt ambitious creative projects.
Example: The interactive portfolio elements I never would have tried before
Insight: Sometimes you need a collaborator who never judges your "stupid" ideas
Skill development: Working with AI has made me better at breaking down complex problems into manageable pieces.
Before: Feeling overwhelmed by big technical challenges
After: Naturally thinking in terms of step-by-step solutions
Surprise discovery: Claude Code fundamentally adapts its coding approach based on the domain you're working in.
The Pattern I Noticed:
Web development: Tends toward production-ready code with complete implementations
Game development: Incremental, placeholder-heavy, TODO-driven prototyping approach
"Completion" means different things: Web = done and polished. Games = working prototype, next chunk unlocked.
The Concrete Example: While building a Minecraft mutualism mod, I discovered leaves were withering away even though they were attached to wood blocks. Bug, right?
Not exactly. Investigation revealed Claude had set up custom blocks using generic Minecraft block models—enough to get them rendering and testable in-game—but not properly inheriting from the specialized PillarBlock and LeavesBlock classes they needed for correct behavior.
It was intentionally halfway. The blocks worked well enough to test core game mechanics, with the expectation of refactoring to proper specs later.
The Insight: This is actually brilliant for game development. Game mechanics need to be felt and tested iteratively to know if they're even fun before you invest in polish. You don't need perfect pathfinding to test if "follow the bird to honey" is an engaging mechanic. You don't need optimized algorithms to validate whether "destroy nest → no fruit" feels right as a game consequence.
Key Lessons:
Embrace domain-appropriate workflows - What works for web doesn't always work for games, and vice versa
TODOs as creative scaffolding - In rapid game prototyping, TODOs aren't tech debt—they're markers of "good enough to test the next idea"
Know when "done" means "done" - Understand the difference between "works for testing" and "production ready"
Expect refactoring - Halfway implementations to enable rapid iteration are a feature, not a bug
Real-world impact: This realization changed how I approach creative prototyping. Instead of fighting the placeholder-heavy approach, I now lean into it during exploration phases—but with clear markers for what needs production-level implementation later.
Read the full story: Building a Minecraft Mutualism Mod Dev Journal
Pitfall: Implementing solutions without understanding them Solution:
Always ask "Can you explain why this approach works?"
Implement incrementally and test each step
Keep learning fundamental concepts alongside AI assistance
Pitfall: AI providing outdated or framework-specific solutions Solution:
Always specify versions and context upfront
Verify suggestions against current documentation
Test in small, isolated environments first
Pitfall: AI makes everything seem possible, leading to overambitious project scope Solution:
Define clear project boundaries before starting
Use AI for feasibility checking, not just implementation
Regular check-ins on project scope and timeline
Pitfall: AI-generated database changes often omit proper migration management Real Example: Working on Django projects where migrations were consistently generated without meaningful names and excluded from commits Solution:
Always explicitly request: "Create a migration named [descriptive_name]"
Run git status after any database changes to verify migrations are included
Establish and communicate clear migration naming conventions
Ask AI to consider cross-app migration dependencies
Pitfall: Moving too fast and not capturing valuable insights Solution:
Keep learning logs (like this post!)
Share discoveries with others
Create reference materials for future projects
Comfort level: Confident using AI for most technical tasks while maintaining healthy skepticism
Workflow: Established reliable patterns for AI-assisted development
Learning focus: Deepening understanding of AI capabilities and limitations
[ ] Expand Creative Applications: Exploring AI assistance in visual design and art projects
[ ] Improve Prompting Techniques: Studying advanced prompt engineering for better results
[ ] Community Engagement: Sharing learnings and learning from others in AI development communities
[ ] Teaching & Mentoring: Helping others navigate AI-assisted development
[ ] Tool Development: Building custom tools that enhance AI workflows
[ ] Content Creation: Regular blog posts documenting ongoing learning
Current Portfolio - Showcasing AI-assisted design and development
Interactive Snake Game - Creative collaboration between human and AI
Artwork Gallery - AI-assisted content management implementation
Symbiotic Survival Minecraft Mod - Game development prototyping with domain-specific AI workflows (dev journal)
He's Going the Distance Minecraft Mod - Fast-iteration gameplay mod development and co-op playtesting
Wagtail Scheduling Version Confusion - Documentation of version-specific challenges
Epic Astro Migration Notes - Lessons from a major framework migration
Performance Optimization Journey - Learning to optimize AI-suggested solutions
Primary AI Assistant: Claude (with Claude Code for development)
Version Control: Git with AI-assisted commit messages and documentation
Documentation: AI-generated technical documentation with human review
Learning Resources: [TO BE ADDED - specific courses, books, communities]
The most important lesson I've learned is that AI is not a replacement for learning - it's an accelerator for learning. The developers who will thrive in this new landscape are those who use AI to deepen their understanding, not to avoid it.
Every challenge documented here, every success celebrated here, is part of a rapidly evolving landscape. What's true today might be different tomorrow, and that's exactly why documenting this journey feels so important.
What's next for this post? I plan to update this regularly as I continue learning. If you're on a similar journey, I'd love to hear your experiences and lessons learned.
This post was created as a living document and will be updated regularly with new insights, projects, and lessons learned. Last updated: October 2025
Connect with me: [Contact information or links to social profiles]
[TO BE ADDED - chronological overview of major AI-assisted projects]
[TO BE ADDED - detailed breakdown of tools used, versions, and compatibility notes]
[TO BE ADDED - successful prompt patterns for different types of tasks]
[TO BE ADDED - common issues and solutions for AI-assisted development]
The songs I liked in January 2025.
![(I'm) Stranded [Deluxe Edition] — The Saints · 1 song](https://i.scdn.co/image/ab67616d0000b273285e8b9d496855b178727df0)
111 songs