Skip to content
Built in AI Skunkworks Growth Rocket's AI Development Lab

Luke · built in Growth Rocket's AI Skunkworks

Catch the broken layoutbefore your client does.

Luke checks your site at every screen width, pinpoints exactly what breaks and where, re-checks after every update, and writes the bug report from what you circle.

License Luke and you get a testing kit maintained and kept current by Growth Rocket: the layout checks, the device list and the screen-size data it runs on.

01Check every width

A few devices show you a few widths. Layouts break in between.

Test on a handful of devices and you've checked a handful of widths; your visitors bring every size in between. Luke renders the page at every width from 320 to 1440 px, 40 px apart, and measures each one. Broken widths are grouped into ranges described in plain words, and any range opens in a live device view with one click.

Range
320 → 1440 px
Step
Every 40 px · 29 widths
Finds
The ranges where the layout breaks, in plain words
Then
One click to a live device view at that width
Luke's Test screen: the Harbor Café page live on an iPhone 15, a Pixel 9 and a custom 320 px device, with the QA toolkit beside themThe Check every width sheet starting: rendering widths, 0 of 29Check every width result: the layout breaks in 1 range, 320 to 400 px scrolls sideways, with the broken thumbnails marked red
320–400 px · scrolls sideways3 of 29 widths · example site 320 px · banner wider than the screen.promo-banner · heading cut off 320 px · scrolls sideways

Devices show a few widths. Luke checks every one in between.

02Mark it, the AI writes the bug

Circle the problem, and the AI writes the report from your marks.

Report from any live device view and mark up the screenshot first: a circle or box, an arrow, a note, the pen, or redact anything private. Then Luke's AI drafts the title, the severity and the description from what you marked, in three parts: what happens, what was expected and where to look. Device, page, viewport, console and network come attached automatically, and Luke warns you when a bug looks already reported. The AI only describes what you marked, and you review and edit the draft before saving.

Mark up
Box, arrow, text, pen, redact · keyboard keys for each
✦ AI draft
Title, severity, what happens, expected, where to look
Attached
Device, page, viewport, console and network
Before saving
A duplicate warning, and your edits
Luke's Test screen with the Report buttons in the QA toolkitReport step 1, Mark up: the Pixel 9 screenshot with the promo banner circled and a text note, and the tools Arrow, Box, Text, Pen and RedactReport step 2, Describe: the marked-up screenshot beside an empty form and the Draft with AI buttonDraft with AI working on the descriptionThe drafted issue: title, what happens, what was expected, where to look, severity Major, and the device, page, viewport and logs attachedIssue #4 savedIssue #4 in Luke: Major, what happens, what was expected, where to look, the marked-up screenshot and the environment
2 marks · a circle and a note AI draft from the 2 marksExample wording for this demo

Mark it up. The AI writes it up. You review it.

03The screens your visitors actually use

Pick devices from the screens people use, then check every width between them.

A device list made from habit misses the screens in between. Luke reads monthly screen-width data (StatCounter), shows how much of it your device set covers, and suggests the smallest set of devices that covers 90% of the screens in use. Save it as a device set; Check every width covers the widths between them.

Luke's Device coverage page: StatCounter screen-width shares for phones worldwide and how much of them the active devices coverSmallest set that covers 90%: four devices reach 94%, each with its width and how much it adds
StatCounter screen widths · October 2026Luke's built-in market data, not your traffic 4 devices reach 94% 4 devices · 4 widthsCheck every width covers 320–1440 px between them

The screens people use, and the widths between them.

04A testing kit Growth Rocket keeps current

You don't just license a testing tool. The kit inside it is maintained and kept current by Growth Rocket.

Anyone who licenses Luke gets the layout checks, the device list and the screen-size data Growth Rocket maintains for its own website work, kept current as new phones, browsers and screen sizes arrive. New devices are picked up by a daily check and approved before they join the library. No off-the-shelf testing cloud comes with an agency's QA practice built in.

  1. 01 Growth Rocket's website work The checks, devices and screen data we rely on for our own client sites, every day
  2. 02 The layout checks Tap targets, small and clipped text, overlaps, fixed overlays, placeholder text, upscaled and broken images, zoom disabled, layout shift, sideways scroll, plus WCAG A/AA
  3. 03 The device list and screen data Phones, tablets, foldables and 26 desktop sizes, a daily check for new devices, and monthly screen-width shares
  4. 04 Your Luke license All of it built in, and maintained and kept current by Growth Rocket
Every check, every device, kept current for you
Luke's device library in the launcher: Popular, iPhone, Android, Foldables, Tablets and Laptops & desktops, each with its devices and sizes
Kept current by Growth Rocketa daily check adds new devices after approval 26 desktop sizes up to 5120 × 1440plus any custom size

The device list, maintained and kept current by Growth Rocket.

A testing kit Growth Rocket keeps current, in your own Luke. Contact Growth Rocket for licensing

05Live device views

Several devices side by side. One scroll moves them all.

Pick a page and devices in the launcher (it tells you what's missing before you open), and Luke opens live device views in browser engines (Chrome's Chromium, Safari's WebKit and Firefox's), each set up with the device's screen, density and touch. Tap, drag, scroll, type and rotate; sync scrolling, taps and typing across devices; switch to dark mode; slow the network on Chromium; and read each device's console and network log.

Luke's launcher: What should we open, and where? A page field and a device libraryThree live device views side by side: iPhone 15 in WebKit, Pixel 9 in Chromium and a custom 320 px device, with sync switched on
Open says what is missing: a page and a device Sync scroll, taps and typing Rotate · dark mode · networkslow the network on Chromium

Live device views, side by side, in browser engines.

06Every page, every device

Every page on every device, and what's wrong on each, measured.

A screenshot run captures your pages on many devices at once, viewport or full page. Then Luke measures every capture: tap targets too small, text too small or clipped, overlaps, fixed overlays, placeholder text left in, upscaled or broken images, zoom disabled, layout shift and sideways scroll, plus WCAG A/AA accessibility checks. Open any shot large, download them all as a ZIP, or open one in a live device view.

The Harbor Café project in Luke: its screenshot runs and width checksA screenshot run on five devices: Galaxy S24, Pixel 9, iPhone SE and iPhone 15, each with its findingsOne capture opened large: numbered boxes on the promo banner and the findings, including Page scrolls sideways, 441 px wide on a 375 px screen, div.promo-banner
441 px wide on a 375 px screendiv.promo-banner · measured, not guessed

Every page on every device, and what's wrong on each.

07Baselines and every deploy

Approve a baseline once. Every deploy is checked against it.

Approve a screenshot as the baseline and Luke compares each new run with it: the screenshot, the differences, the baseline or a slider between them, with every changed area boxed and graded major, minor or cosmetic. Luke's AI sorts each change into intended, regression or noise and explains it in a sentence.

A run where every device is the approved baselineThe next run: 5 changed since the baseline, each device 13 or 14% differentDifferences: the changed pixels in red, grouped into 2 changed areas with their sizesThe slider between the baseline and this runThe findings for the change, including the AI note that it looks like a bug
Sorted as a bug, not an intended changeExample wording for this demo

What changed since the approved baseline, and whether it should have.

Runs on a schedule, or after every deploy.

Set a schedule, or point your deploy at Luke's webhook. Each run is compared with the last, and the people watching the project hear only about new issues.

The Harbor Café project overview in Luke: pages, recent activity, the set-up checklist with scheduled screenshots and a deploy hook, and the latest issue
Watch · hear about new issues only

On a schedule, or after every deploy. Watchers hear about new issues.

08One link, two audiences

One link, two audiences: a clean page for your client, the details for your developer.

Share a run, a width check or a set of issues as a clean, read-only page for your client, or a developer page with the technical details, with an expiry date you set: 7, 30 or 90 days, or until you stop it. See how many times it was viewed and send it in a branded email. Bugs also export as Markdown, a Jira-ready CSV, a PDF, a ZIP with the images, or an email with the PDF attached.

A screenshot run in Luke with the Share buttonShare these screenshots: who is it for, a client or developers, and how long the link worksThe share link: client view, expires 20 Dec, opened 0 times, with Email it and Stop sharingThe client's read-only page: Luke by Growth Rocket Propulsion Labs, Harbor Café, 1 page on 5 devices, each device with its sizeLuke's Issues page: export to Jira (CSV), Markdown, PDF or a ZIP with images, or share and email them
Client view or developer details 7, 30 or 90 days, or until you stop itwith a view count The same share page on a client's phone

A clean page for your client, the details for your developer.

09A second pair of eyes

An AI second look at every run, next to what Luke measures.

AIAsk for an AI review and Luke's AI looks over up to 24 screenshots from a run, writes a summary and explains what changed against the baseline. Luke labels it as a model's view to verify before acting. Alongside it, Insights counts the findings that keep coming back across runs and sites, from the measured checks only, so the numbers can be reproduced.

A screenshot run with the AI review button and the AI summary card above the devicesLuke's Insights: Keeps coming back, overflow-x on div.promo-banner in 2 runs on 2 devices; trend per project; devices with the most findings
An AI second look: verify before actingExample wording for this demo Plain counts · the same rule on the same selectorAI observations are left out, so the numbers can be reproduced

An AI second look at every run, alongside what Luke measures.

10The rest, briefly

Everything else a testing team needs, already in.

  1. Staging login in the project set-up checklistSave login on each live device view

    Staging, logged in once

    Save a staging login from the live view, or set basic auth, headers or cookies per project.

  2. A Luke project: Overview, Screenshots, Issues and Settings, with pages, recent activity and the set-up checklist

    Projects

    Each site is a project with Overview, Screenshots, Issues and Settings, its pages, recent activity and a set-up checklist.

  3. Luke's Issues: filters by status and severity, and the filed issue with its severity, device and project

    Issues in one place

    Every issue across projects, filtered by status and severity, with its device and project.

  4. An issue in Luke with its environment: device, viewport, user agent, page and when it was captured

    Every issue carries its environment

    Device, viewport, user agent, page and the moment it was captured, attached automatically.

  5. The full width report: the layout breaks at 3 of 29 widths, 320 to 400 px, clean again from 440 px up, with the striped bar and thumbnails

    The full width report

    Every width check opens as a page: the broken range, where it is clean again, and all 29 widths.

  6. Luke's sidebar: Test a site, and Search with the ⌘K shortcut

    Search, from the keyboard

    Press ⌘K to search Luke from anywhere.

That's the whole testing hub, in your hands. Contact Growth Rocket for licensing

Find it, write it up, hand it off.

Luke is the testing hub Growth Rocket uses on its own client sites, with a testing kit maintained and kept current by Growth Rocket. Talk to us about licensing it.

Contact Growth Rocket for licensing

What buyers ask

Does Luke test on real devices?

Not on physical phones, and we say so. Luke renders in browser engines (Chrome's Chromium, Safari's WebKit and Firefox's) set up as each device's screen, density and touch. It's built to catch layout and responsive bugs early and on every deploy; quirks of one physical phone still need a check on that phone.

Why not just use BrowserStack?

Use it for physical-device checks. Luke does the part a device cloud doesn't: it measures every width, finds the break, and writes up the bug from your marks.

Can AI-written bug reports be trusted?

The AI only describes what the tester marked, and the tester reviews and edits the draft before saving it. The run review is an AI second look, not a verdict.

How does it fit our developers' workflow?

Bugs export as a Jira-ready CSV, PDF, Markdown or a ZIP with images, or go out as a link with technical details. Checks can run on a schedule or after every deploy.

What do clients see?

A clean, read-only page without the technical noise, with an expiry date you set.

What does 'kept current' cover?

The layout checks, the device list (with a daily check for new devices) and the screen-size data behind coverage are maintained and kept current by Growth Rocket, for its own work and for every licensee.

Can we buy Luke?

Luke is Growth Rocket's own tool, used in our website work. Talk to us about licensing it.

Luke was built in Growth Rocket's AI Skunkworks.

AI Skunkworks is Growth Rocket's AI Development Lab: we build AI-powered applications that solve real business and operations problems. Live device views in three browser engines, every-width checks, visual baselines, scheduled runs, AI bug write-ups and client-ready sharing: Luke is a complete testing platform, designed, built and run by Growth Rocket. We use it every day on our own client sites, and the AI Skunkworks team maintains it and keeps its testing kit current. When the right tool doesn't exist, we build it.

Build an AI app with Growth Rocket
  1. 01 Cover
  2. 02 Check
  3. 03 Find
  4. 04 Share

About Growth Rocket

Growth Rocket is a full-service growth marketing agency for SEO, GEO and AEO, paid media, and web and AI application development, with offices in Los Angeles and Manila. Its AI Skunkworks lab builds the AI tools its own teams run on.

Visit growth-rocket.com
Work With Us