From Tinkerer to Builder: How Claude Code Changed What I Think Is Possible
It happened somewhere between the third coffee website iteration and realizing I was building a personal media tracker at 10 PM on a Tuesday. Yes, my ADHD brain flounders and spins between so many projects it is miracle I ever complete anything.
I looked at my screen and had one of those moments where you see yourself from the outside. Three weeks ago, I would have added “build a website” to my someday list of 2026. Today, I had five of them running. Not someday. Today.
This isn’t a story about learning to code. This is a story about what happens when the gap between “I wish this existed” and “I built this” shrinks from months to hours.
The Projects (Or: What Happens When You Stop Punting)
In the past few weeks, I’ve built more than I built in the previous year. Not because I suddenly learned Python or became a developer. Because Claude Code let me stay in the problem space instead of drowning in the solution space.
Here’s what actually got made:
The Tool Vetting Spreadsheet Saga Remember my recent post about flow state and not wanting to interrupt colleagues? This was patient zero. I had a vision for a spreadsheet that would automate district AI tool evaluations using the VATT Framework. I needed complex formulas, conditional formatting, and logic that was frankly above my pay grade. Instead of punting or waiting for the next professional development day when someone could help me, I sat down with Claude Code. We iterated. We broke things. We fixed things. The spreadsheet now does exactly what I envisioned - and I learned more about structured thinking in those sessions than I would have in any tutorial.
The Student AI Website Evolution This one started as a joke. I wanted to create an AI explainer website using Simpsons characters because, well, that’s how my brain works. Then I realized the copyright issues. Then I pivoted to a coffee shop metaphor. Then it became AI Career Brew: Your Future Ready Menu, an actual interactive website where students learn about AI concepts through a barista simulator and data storytelling.
https://futurereadyai.netlify.app/
The project didn’t just change mediums. It changed three times in concept. In the old world, that kind of pivoting would have meant starting over, new mockups, new development cycles. With Claude Code, pivoting meant continuing the conversation. “Actually, let’s make this about coffee instead” wasn’t a restart; it was the next iteration.
The NASA LEGOnauts Project Site I’ve helped facilitate the LEGOnauts program at a local school connecting students with real NASA and computer science and builders who are experts for authentic space exploration challenges. But the program lived in scattered Google Docs and slides. It deserved a home. So I built one. A website that captures the philosophy, showcases student work, and helps other educators understand what authentic expert connection looks like in STEM education.
The site isn’t just documentation. It’s a recruitment tool, a philosophy statement, and a proof of concept all in one.
It is not yet ready as I am still editing video and so forth, but when ready to launch I will share.
The Mall: My Personal Media Tracker This one is pure self-interest, and I’m not apologizing for it. I read a lot. I watch a lot. I wanted a place to track books, shows, thoughts, and recommendations without feeding more data to corporate recommendation engines. So I built “The Mall” of my 80’s youth for my personal media tracking website.
It’s not for anyone else. It’s for me. And that’s the point. When building becomes accessible, you can make things just because they’d be useful or delightful to you personally. Not everything needs to scale or serve a broader mission.
What Actually Changed (Hint: It Wasn’t Learning Syntax)
I didn’t learn to code in these three weeks. I learned something more fundamental: I learned to think in iterations instead of destinations.
Old Mindset:
I need a website
I don’t know how to build websites
I should take a course / watch tutorials / find someone who knows
Maybe next semester
New Mindset:
I have an idea
Let me describe what I want
Let’s build a rough version
That’s not quite right, let me adjust
Actually, what if we tried this instead
Oh, that works better
Ship it
The shift isn’t technical. It’s philosophical. I’m no longer asking “Do I have the skills to build this?” I’m asking “Can I describe this clearly enough?”
The Navigation Part (Or: What I Actually Did)
When people ask me how I “learned” Claude Code, I tell them I didn’t learn anything in the traditional sense. I navigated. Here’s what that looked like:
Started with something I actually needed: Not a tutorial project. Not a practice exercise. The training materials I was already building. Real stakes, real deadline, real purpose.
Embraced terrible first drafts: That Simpsons website? Objectively bad idea from a copyright standpoint. But it got me started. It got me past the blank page. Every project started messy and got refined through conversation.
Learned to describe outcomes, not methods: I stopped saying “I need a Python script that does X.” I started saying “I need this spreadsheet to automatically flag tools that don’t meet FERPA compliance based on these criteria.” Claude Code figured out the how. I stayed focused on the what and why.
Iterated in public (with myself): I kept a running commentary of what was working and what wasn’t. Not for anyone else. For me. Writing “This feels clunky because...” helped me articulate what needed to change.
Built everything twice: First version was always exploratory. Second version was always “Now that I know what this should actually be...” The cost of iteration is so low that rebuilding became a feature, not a bug.
The List (For Those Who Want the Receipts)
Projects created in the past three weeks using Claude Code:
VATT Framework evaluation spreadsheet with automated scoring
Tool vetting assessment templates
Coffee For The Brain student AI education website (barista simulator version)
NASA LEGOnauts program website
The Mall personal media tracker
And soon to be shared projects in development
None of these are perfect. All of them are functional. Most importantly, all of them exist because I stopped waiting for permission or expertise and started building.
The Conversations We’re Not Having (Or: Why This Isn’t Really About Claude Code)
Here’s what keeps happening in my workshops and district meetings: We start talking about AI tools. Someone shares a study. Someone else shares a concern. And before long, we’re not actually talking about AI at all.
We’re talking about control. About trust. About what learning actually means.
A recent study landed in my inbox: learning via ChatGPT produces shallower knowledge than learning via Google Search. It’s well-designed research with over 4,500 participants. The kind of study that gets cited in board meetings and policy documents.
A decade ago, research showed that Google Search led to worse retention than print materials. The pattern is striking: Print > Web > AI. Each new technology becomes the existential threat while the previous one gets quietly rehabilitated.
Punya Mishra captures this perfectly in his post “The Nostalgia Machine: Why Ed-Tech Research Keeps Missing the Point”. We’re not discovering that new tools make learning worse. We’re discovering our own nostalgia bias playing out in laboratory conditions.
And Gary Stager nails the absurdity: the same people who were terrified of Wikipedia a few years ago are now using it as the gold standard to prove ChatGPT is dangerous. The cyberostriches, as he calls them, always have their heads in the sand about the current moment while romanticizing the tool they previously feared.
So what’s actually happening here?
Is it nostalgia bias? Absolutely. The tool we grew up with always feels more “authentic” than the new one.
Is it a habituation problem? Probably. We’ve had fifteen years to develop effective search strategies. We’ve had two years with AI chat interfaces.
But here’s what bothers me most: It’s a framing issue.
When we operationalize “learning” as memorizing trivia, we’re measuring precisely the kind of knowledge that external memory systems render less necessary. We’re testing for the skills that the previous generation of tools already made obsolete while wringing our hands about the new generation.
Meanwhile, while we conduct careful laboratory studies comparing reading comprehension across platforms, recommendation algorithms are reshaping young people’s relationships to truth, identity, and institutional authority. The transformations that actually matter to democracy and identity formation remain largely outside our methodological reach.
We’re arguing about whether ChatGPT is Wikipedia while completely missing the question of what we’re trying to build in the first place.
What This Means for Educators (The Part Where I Get Philosophical)
I work in education. I help districts navigate AI policy and implementation among many other things. And I’m increasingly convinced that the most important skill we can teach isn’t prompt engineering or AI literacy in the traditional sense.
It’s the ability to move from consumer to creator with nothing but clear thinking and iteration.
And it’s the ability to ask better questions than “Is this tool good or bad?”
Because here’s what my three weeks of building taught me: The conversation about whether Claude Code (or ChatGPT, or whatever comes next) makes us “shallower” thinkers misses the point entirely.
The question isn’t whether the tool makes us better or worse at memorizing information. The question is what becomes possible when we’re no longer blocked by technical implementation.
When the gap between idea and implementation shrinks, we stop asking permission. We stop waiting for the expert. We stop punting when we hit a wall.
We build the thing we wished existed.
And that changes the questions we should be asking:
Not “Did students retain the facts?” but “What did they create that didn’t exist before?”
Not “Is this making them dependent?” but “What are they now capable of that seemed impossible last month?”
Not “Is this tool safe?” but “What happens when building becomes accessible to everyone?”
For educators specifically, this matters because:
We’re drowning in tools that almost work: How many times have you used a platform that does 80% of what you need? With Claude Code, that remaining 20% becomes buildable. Not by IT. By you.
We have domain expertise that matters more than technical expertise: I know what makes good AI training materials. I know what districts need to evaluate tools effectively. I know what engages middle school students in STEM. Those insights are more valuable than knowing JavaScript.
Our students need to see us as builders: If we want students to see themselves as creators, we need to model that. Not by becoming developers, but by showing them what’s possible when you combine clear thinking with powerful tools.
The Uncomfortable Truth
Here’s what I’m still wrestling with: This feels almost too easy.
Not the building part as that still requires effort, iteration, and clear thinking. But the barrier to entry. Three weeks ago, building a website felt like a semester-long project. Today it feels like an afternoon.
That’s exciting and unsettling in equal measure.
Because if I can do this and someone who took exactly one programming class in college and promptly forgot everything, then the question isn’t “Who can build?” anymore.
The question is “What should we build?”
And maybe that’s the real anxiety underneath all those studies comparing print to web to AI. It’s not that each new tool makes us shallower. It’s that each new tool forces us to reckon with what depth actually means.
When memorization becomes less necessary, we have to figure out what thinking actually is.
When technical barriers disappear, we have to decide what’s worth creating.
When anyone can build, we have to ask what we’re building toward.
Those are much harder questions than “Is this tool good or bad?”
And I don’t think we have good answers yet.
What’s Next
I’m going to keep building. Mostly because I can’t stop at this point. The next project is already forming.
Will I “know how” to build it in the traditional sense? Probably not.
Will I build it anyway? Absolutely.
Because the old rules about who gets to create things don’t apply anymore.
And that changes everything.
What have you been punting on because you didn’t have the “right” skills? What would you build if the gap between idea and implementation shrunk to hours instead of months? Hit reply - I’d genuinely love to hear what’s on your someday list.