Browser.ai
Blog

CBrowser Blog

Articles on browser automation, cognitive testing, AI safety, and the future of web quality.

We Beat Two Frontier Vision Models With Arithmetic
8 min
researchattentionbenchmarks+2

We Beat Two Frontier Vision Models With Arithmetic

Our attention model lost to a blurred circle in the morning and beat GPT-5.6 Sol and Claude Fable-5 by the afternoon. Three unglamorous fixes did it: a constant nobody had measured, a property of the eye nobody had modeled, and a standard technique nobody had implemented.

Alexandria Eden
Your Mega Menu Is the Hardest Thing on Your Site
7 min
uxaccessibilitynavigation+1

Your Mega Menu Is the Hardest Thing on Your Site

180 links, invisible above the fold, and the biggest cost on the page for the two visitors with the least to spare.

Alexandria Eden
What Cognitive Effort Actually Measures
10 min
cognitive-effortaccessibilityux+1

What Cognitive Effort Actually Measures

Your performance tools measure the machine. Your accessibility tools measure the markup. Neither measures the person, and the person is the one who leaves.

Alexandria Eden
Where a Simulated Person's Motivations Come From
13 min
engineeringpersonasresearch

Where a Simulated Person's Motivations Come From

A persona in cbrowser is thirteen numbers describing what a simulated user wants. There are three ways to get them, they are not equally trustworthy, and the output tells you which one you are holding.

Alexandria Eden
Screen Capture: Recordings Your AI Can Actually Read
4 min
releasescreen-capturerecording

Screen Capture: Recordings Your AI Can Actually Read

CBrowser 18.72 records any page to GIF, WebP, or WebM β€” and emits a frame manifest with timestamps, change points, and per-frame console/network events, so an AI agent can answer "what happened when" without watching a video.

Alexandria Eden
What We Learned About Readability
5 min

What We Learned About Readability

We ran seven languages across four script families through CBrowser's cognitive assessment. German compound words demand more working memory. CJK carries more meaning per character. Arabic adds RTL cognitive load. Two tokenization bugs caught and fixed in one session.

Alexandria Eden
A Centered Blob Beat Our Attention Model - What 495 Pages of Eye-Tracking Showed
12 min

A Centered Blob Beat Our Attention Model - What 495 Pages of Eye-Tracking Showed

We re-ran our saliency benchmark against the code we ship, scored on real fixations from 62 people. Center bias beats every persona on 78% of pages. The persona filters change the heatmap 4.5x and its accuracy by under 0.005 - they redistribute attention without predicting it.

Alexandria Eden
500 Free Credits That Never Expire β€” Our Pro Launch Promotion
3 min

500 Free Credits That Never Expire β€” Our Pro Launch Promotion

Sign up for CBrowser Pro and get 500 bonus credits on top of your 2,000 monthly allowance. They never expire. Here's what you can do with them.

Alexandria Eden
Why Accessibility Checkers Can't Tell You Who Will Struggle β€” And What Can
8 min

Why Accessibility Checkers Can't Tell You Who Will Struggle β€” And What Can

Cognitive Optimal Transport measures the gap between what users' brains can handle and what websites demand β€” predicting who will struggle before they ever visit.

Alexandria Eden

Showing 9 of 0 posts