Google DeepMind
Antigravity
A dark, interactive first iteration with a circuit-canvas motif and a technology-forward presentation of the portfolio.
Portfolio experimentEdition 01—03
Same brief.
Different builders.
A public, real-world comparison of how AI coding assistants interpret one portfolio brief—through architecture, visual judgment, content handling, and day-to-day workflow.
03Assistants
01Shared brief
RealProject constraints
PublicBuild notes
01 / Iterations
Three approaches.
Kept deliberately separate.
Each version receives the same factual source material. The implementation, aesthetic, and engineering choices are allowed to diverge.
OpenAI
Codex
Silicon design, engineering case studies, and an evolving collection of hardware and software projects. Explore the portfolio and its hands-on AI experiments.
Anthropic
Claude Code
The third portfolio iteration is reserved for a future build and comparison.
02 / Method
Not a synthetic benchmark.
The comparison happens inside a maintained, published website. Every assistant has to work with existing content, inherited constraints, accessibility needs, and real deployment.
- 01Source
Begin from the same neutral professional and blog content.
- 02Build
Let each assistant establish its own architecture and visual language.
- 03Observe
Review clarity, maintainability, interaction, accessibility, and accuracy.
- 04Document
Keep the versions public so the differences can be explored directly.
Current focus / OpenAI Codex