"If you can dream it, you can build it..."
Marketers have certainly gotten a piece of the AI-boom's pie over the last few months, and, maybe for the first time, they're grandiose statements are matched by uncannily good products.
I've put my foot down when it comes to using LLMs for the written word and image generation. The dangers in these areas of innovation are abundantly clear. Either the skills we have in these spaces are too valuable to lose or the cost of employing generative superpowers at scale is too dangerous to unleash societally.
The area of coding (and the abilities Claude's Opus models have gained over the last 3 months with artifact-creation, batch document editing, and workflow optimization) has been one I have yet to unpack fully. Until now...
I've spent the last thirty days deep inside Opus 4.5 and, since its middle of the month release, Opus 4.6.
At this point, the hype is ubiquitous. People have flooded to this new competitor to OpenAI's models whose formerly beloved chatbots are becoming simultaneously more sycophantic and less capable in Claude's shadow. Last week after geopolitical moves effected public perception of these company's ethics and abilities, Claude went down due to the scale of migration from ChatGPT.
The promise, "If you can dream it, you can built it", has been borne out. You can build it.
With no coding experience, I have a 27,000 line HTML file in a root directory with more than 500 transitions and animations, 15 images and SVG icons, "_redirects" from an old site, custom font folders, and a sitemap for Google indexing.
And I have only vaguely know what any of that means.
My website, formerly hosted by Squarespace for $260 a year, is everything I've ever wanted a website to be at $20 a year.
Inspired by organizations, artists, concepts, and avant-garde styles I've collected in bookmarks over the years, I finally have a space where I am proud for my writing, ideas, and practices to live. It's shape and systems may be a mystery to me on the backend, but what is present visually in the user experience and interface on the front end has become the stuff of my month-long dreams. Go check it out.
In the midst of limitless creative capacity, I encountered the primary peril of AI's application in this novel (for me) space: I could have spent an unlimited number of hours fervently building with Claude and never looked up.
To their credit, Anthropic has imposed a very easy to reach "usage limit" for those working with their advanced capabilities (of which website development is currently one; but our current rate of progress indicates this won't be so "advanced" for long) that can be increased with more expensive plans or waited out in 6-12 hours.
As I dreamt, I built.
The promise was fulfilled, but questions remain... When will I be finished dreaming? When is enough enough? What does it mean for me to have built this ex nihilo (both creatively and skillfully)? Does the creation of something beyond my understanding and capacity limit my inhabiting of it? How does the reification of a collaborator's ideating and artificing my site change my way of relating to its future outputs, even beyond that collaborator's contribution? And (a big one) have I chained myself to a new and virulent technology in the population of my domain with its code, its structures, and its execution of my wants and needs?
I have deleted Claude from my computer. I no longer have it bookmarked.
My website will remain an artifact of our partnership from this month, but whether the above queries or my site's future needs or the ever-changing nature of life online will make our paths cross again will be seen and encountered as either a benign part of upkeep or a toxic slide back into these question's dark sides and dangers...