Detailed Analysis
A product designer with self-described "basic HTML and CSS skills" built a fully interactive geological clock — eona.earth — that maps Earth's 4.5 billion year history onto a 12-hour timescale, using Anthropic's Claude Code as the primary development engine. The project renders a WebGL globe with paleogeographic continent data, procedural cloud systems, and atmospheric haze that evolve dynamically across 14 distinct phases of Earth's history. Key geological milestones are mapped to specific times on the clock face: the Moon's formation at 12:06, first life at 2:45, the extinction of the dinosaurs at 11:39, and the entirety of human existence compressed into the final three seconds. The result is a single self-contained HTML file with no build step — a deliberate constraint that kept the project tractable for a non-engineer working in short evening sessions over approximately two weeks.
The development workflow reveals a sophisticated and intentional approach to working within the constraints of AI-assisted coding. The creator initially ran Opus 4.7 for the project's first evening but pulled back after finding it over-engineered solutions — one instance had a fragment shader executing four noise passes per pixel at 60fps, taxing devices unnecessarily. The subsequent strategy involved alternating between Claude Sonnet for standard implementation work and Opus 4.6 for more complex or stuck situations, while using Gemini's free Thinking mode to establish a color system that flexed coherently across the geological phases. This multi-model orchestration, with different AI systems assigned to tasks matching their cost-performance profiles, reflects a pragmatic and increasingly common pattern among technically adjacent creators who are learning to treat AI models as a tiered toolkit rather than a single monolithic resource.
The project's most technically demanding element — animating continental drift through deep time — was solved through shape-tweening across paleogeographic continent maps, with Opus 4.6 independently building an interactive palette editor as an unprompted addition. That moment of unexpected capability is significant: the model identified a useful auxiliary tool, built it without being asked, and delivered something the creator described as "very impressive." It illustrates how current-generation AI coding assistants can move beyond mechanical instruction-following into proactive problem-solving within a defined project scope, blurring the line between tool and collaborator. The creator's management of context files and off-peak usage also highlights the real-world friction of working within subscription usage limits on creative personal projects, a constraint that shaped the pacing and structure of the entire build.
Eona.earth sits within a broader trend of non-engineers using AI coding tools to produce technically sophisticated, production-quality interactive experiences that would previously have required a dedicated developer or a team. The project's ambition — custom shaders, geologic data integration, procedural rendering — would represent months of conventional learning for someone without a programming background. With Claude Code compressing that barrier, a product designer was able to ship a working, publicly accessible tool in two weeks of part-time effort. Planned extensions — including a physical build on a Waveshare round display and Raspberry Pi 4, watch and mobile apps, and sound design — suggest the creator views the initial release as infrastructure rather than a finished artifact, an orientation that itself mirrors how software teams approach living products. The project's framing around "deep time" as a source of psychological perspective adds a contemplative dimension rarely found in technical side projects, suggesting that AI-enabled creative tooling is enabling a new genre of work that blends engineering, design, and humanistic intent in ways that resist easy categorization.
Read original article →