Claude Code versus Anti-Gravity: inspect three different build tasks

Keep the working setup attached to the result

The April comparison puts Claude Code and Anti-Gravity through three challenges: mobile app design with Sleek, a landing page and backend API code with Supabase authentication. Setup and model selection appear at 1:09, followed by pricing and rate limits at 2:40. Those details matter when interpreting the recorded outputs; the tools, models and limits are not interchangeable inputs.

The original design resource is Sleek.design: http://sleek.design/?ref=luke. The video introduces skills at 3:24 and Sleek at 4:50 before running the first design task. A skill supplies task-specific guidance, so keep its use visible when comparing results.

1. Write the same task requirements for both routes.
2. Record the model and any skill used for each attempt.
3. Keep the initial outputs before making follow-up changes.

Compare the two design jobs

The mobile app test starts at 5:50, with results at 7:15 and a closer look at providing real context at 8:10. Use the intended app and audience to assess the designs. An attractive screen still needs to express the right content and interaction.

The landing-page test begins at 10:00 and its comparison at 11:29. Treat that as a separate job: the hierarchy and visitor action of a landing page are different from a mobile app screen.

1. Compare the mobile results with the supplied app context.
2. Inspect the landing pages side by side for clarity and the intended visitor action.
3. Note which improvements came from richer context or skill guidance.

Inspect the backend independently

The Supabase API task begins at 12:52 and the functionality test at 14:15. Inspect the requested authenticated behaviour and the actual test result, not just the generated code’s appearance. Keep a failed attempt distinct from a repaired one.

The final chapter at 16:00 names Claude Code as the winner. That is the recording’s verdict on these tasks; the description does not provide a universal model ranking or a complete security audit of the backend. Use the visible functionality evidence alongside the design comparison when considering the workflow.

The takeaway

Compare equivalent briefs and retain the role of skills and context. Then assess backend behaviour separately from design preference.