[BUG] Claude Code public product page has reproducible ~7.3s mobile LCP

Preflight Checklist

• [x] I have searched existing issues and this hasn't been reported yet
• [x] This is a single bug report (please file separate reports for different bugs)
• [x] I am using the latest version of Claude Code

What's Wrong?

Support reference

Anthropic Support conversation ID: 215475144816317

Support suggested reporting the Claude Code-specific finding in this repository.

Affected public page

https://claude.com/product/claude-code

What was observed

A passive three-run Lighthouse experiment using DevTools throttling produced:

• Valid LCP measurements: 3/3
• Median Performance score: 35
• Median LCP: 7.299 seconds
• Median TBT: 2.294 seconds
• LCP range: 6.463–7.588 seconds
• Runtime errors: 0
• NO_LCP results: 0/3

Test profile

• Mobile viewport: 390 × 844
• DPR: 1
• DevTools throttling
• RTT: 150 ms
• Throughput: 1638.4 Kbps
• CPU slowdown: 4×
• Three sequential passive navigations
• No authentication or user interaction

Evidence

Exact report and reproducibility details:

https://github.com/safal207/LiminalQAengineer/pull/68

Vendor-facing summary:

https://github.com/safal207/LiminalQAengineer/blob/agent/claude-code-exact-rerun-v0-1/docs/audits/CLAUDE_VENDOR_SUBMISSION.md

Evidence boundary

This is laboratory evidence, not field telemetry.

The audit did not use authentication, prompts, model/API calls, uploads, fuzzing, exploitation, or load testing.

A previous Lighthouse report returned NO_LCP; its displayed Performance score of zero was not treated as a valid product score. The new finding is based only on the repeated 3/3 valid measurements above.

What Should Happen?

The Claude Code public product page should consistently render its primary content promptly under the same mobile test profile.

Across repeated runs, Lighthouse should report a valid LCP with substantially lower load time and reduced main-thread blocking, rather than the current median LCP of 7.299 seconds and median TBT of 2.294 seconds. The page should remain visually stable, usable, and responsive while loading.

Error Messages/Logs

Steps to Reproduce

1. Open a clean terminal session on a machine with Node.js and Chrome/Chromium installed.

1. Create a Lighthouse CI configuration targeting:

https://claude.com/product/claude-code

1. Configure the test with:

Mobile viewport: 390 × 844

Device scale factor: 1

DevTools throttling

RTT: 150 ms

Throughput: 1638.4 Kbps

CPU slowdown: 4×

Fresh Lighthouse context for each run

1. Run three sequential passive navigations using Lighthouse CI:

npx --yes @lhci/cli@0.15.1 autorun

1. Do not log in, click page elements, submit prompts, or perform any other interaction.

1. Open the three generated Lighthouse reports and record Performance, LCP, and TBT.

1. Observe that all three runs produce valid LCP measurements, with:

Median LCP: approximately 7.299 seconds

Median TBT: approximately 2.294 seconds

Median Performance score: 35

LCP range: approximately 6.463–7.588 seconds

Full reproducible configuration and evidence:

https://github.com/safal207/LiminalQAengineer/pull/68

Claude Model

None

Is this a regression?

Yes, this worked in a previous version

Last Working Version

_No response_

Claude Code Version

N/A — this report concerns the public Claude Code product page at https://claude.com/product/claude-code. No Claude Code CLI or desktop client was installed or executed.

Platform

Other

Operating System

Ubuntu/Debian Linux

Terminal/Shell

Terminal.app (macOS)

Additional Information

_No response_