AIToday
Large Language ModelsAI Coding AssistantsZenn AI/MLPublished: Nov 20, 2026, 04:00 JST

Modern Web Guidance cuts Claude JS from 4,376 to 2,740 lines

Send on LINE
Modern Web Guidance cuts Claude JS from 4,376 to 2,740 lines

3 Key Points

  1. What happened

    With Google Chrome's Modern Web Guidance plugin loaded, Claude Opus 5.5 wrote 2,740 JavaScript lines across 16 tasks versus 4,376 without it, using twice as many modern browser features (4.6 to 8.6 per build).

  2. Why it matters

    The change is less about better-looking output than about approach — delegating work to built-in browser features instead of hand-written script — and it came at a cost of 24% more tokens and 38% more time.

Not sure about something? Ask the AI

Questions and answers are published on this page.

Summaries like this, every morning.

Context & Analysis

The test compared Claude Opus 5.5 with and without the Modern Web Guidance plugin across 16 front-end tasks, from modals and tooltips to CSV aggregation and theme switching. The headline result was a 37% drop in JavaScript lines, but the bigger shift was in approach: with the plugin, the model more often replaced hand-written JavaScript with native HTML and CSS features. For example, a custom select went from 429 lines of JavaScript to 12 using appearance: base-select, and an FAQ accordion fell from 136 to 10 lines. The plugin also nearly doubled the variety of modern browser features used per build, introducing APIs like closedby, interestfor, and scroll-state queries that the model never chose on its own. The trade-off was higher cost: 24% more tokens and 38% longer median task time. There were also failures — one task broke at runtime in Chrome because the model tested only a polyfill path and missed a native initialization bug, and several builds added CDN-loaded polyfills pinned at @latest. The body notes these results are from a single run per condition and may not generalize.

FAQ
How much did the plugin change JavaScript output?
Across 16 tasks, JavaScript lines fell from 4,376 to 2,740, a 37% reduction. In one extreme case, a custom select dropped from 429 to 12 lines using appearance: base-select.
Did the plugin make the code more modern?
Yes, the average number of modern features per build rose from 4.6 to 8.6. Features like closedby, interestfor, popover="hint", and animation-timeline appeared with the plugin but not without it.
What were the downsides of using the plugin?
Token consumption rose 24% and median task time rose 38%, from 134 to 185 seconds. One build failed at runtime in Chrome due to a Temporal initialization bug that only appeared on the native path.
Did the plugin improve visual quality or basic behavior?
Almost no difference was observed in appearance or basic behavior. Opus 5.5 without the plugin already used modern patterns like <dialog> and <details name>, and all behavior tests passed except one runtime error with the plugin.
Send on LINE

AI news that matters for your work, delivered every morning.

Pick your industry and the AI tools you use, and get news related to your work every day.

Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →

Ask AI

Ask AI anything about this article. The AI reads this article, earlier AIToday articles, and Wikipedia, and cites its sources. Q&As are published on this page for other readers too.

Questions and answers are published on this page.

Related Articles

Next articleGoogle's 'Mirror Dimension' trains Chrome AI agents on a fake web