Specialized AI agent Mano-CUA outperforms Claude and Gemini on browser automation benchmark

·1 views

A specialized vision-based AI agent called Mano-CUA 1.1 scored 41.7 on WebRetriever Protocol I, a benchmark testing real-world browser navigation and form-filling tasks. By comparison, Gemini 2.5 Pro Computer Use scored 40.9 and Claude 4.5 Computer Use scored 31.3, highlighting a notable performance gap. Unlike general-purpose models that rely on DOM parsing or accessibility APIs, Mano-P uses a pure vision-driven approach, interpreting screenshots the way a human would rather than reading page structure. This method proves more resilient on complex, dynamic web applications where page layouts shift after each user interaction. The model can run entirely on-device on Apple Silicon hardware with 32GB RAM, keeping user data local without sending screenshots to external servers.

Read the full story at DEV Community

This is an AI-generated summary. ShortSingh links to the original source for the complete article.

Discussion (0)

Tradesired WebTrader Review: How Its Browser Client Handles Real Trading Demands

A technical review of Tradesired's WebTrader platform examines how the browser-based trading client addresses the core engineering challenges of web trading, including live data streams, chart rendering, and order state management. Unlike native desktop clients, browser trading tools must operate within the constraints of a single tab, making reliable socket connections and accurate state restoration especially difficult. The platform's interface organises instrument navigation, order entry, and charting into a familiar layout, with small details like four live session clocks for major financial centres noted as practical additions. The review highlights the advantage of web-based trading apps in keeping all users on the latest version, eliminating a category of support issues tied to outdated desktop builds. State management is identified as a key design challenge, with the distinction between server-side financial data and local UI preferences requiring different storage and restoration strategies.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Tutorial: Build a Fed-focused Polymarket Research Brief Using Python and AlphAI

A developer tutorial on DEV Community demonstrates how to automate Polymarket prediction-market research using Python 3.10 and the AlphAI financial news API. The script pulls market data from Polymarket's public Gamma API, combining it with AlphAI's macro news feed and economic calendar to generate a structured JSON research brief. The example focuses on a Federal Reserve rate-decision market, gathering relevant inflation coverage, FOMC dates, and settlement rules in a single output file. AlphAI's free tier allows up to 100 API requests per day without requiring a credit card, making the setup accessible for casual use. The guide is intended for researchers or developers who want to consolidate prediction-market context without manually browsing multiple sources.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

How Munchable Solved Design Token Drift Across App, Reels, and Marketing Site

Munchable, a product with surfaces spanning an Expo app, Next.js marketing site, and Remotion social reels, faced a recurring problem where hand-copied design tokens drifted out of sync across platforms. A change to a colour or label in the app would leave the video reels rendering outdated values, with no type error to flag the mismatch. To fix this, the platform-agnostic tokens — including colours with contrast ratios, spacing, radius scales, and verdict labels — were moved into a shared workspace package consumed by both the app and the reels. Each consumer then adds only its platform-specific needs, such as React Native font names for the app or CSS font stacks for the reels. The marketing site was deliberately excluded from the package because its design intentionally differs, using larger corner radii suited to a landing page rather than a compact mobile UI.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

CSS gains inline conditionals and custom functions in Values & Units spec

The CSS Values & Units specification is introducing two new features: the if() conditional function and reusable custom functions that accept arguments. The if() construct allows developers to select property values inline based on media, supports, or style conditions, without duplicating selectors across multiple at-rules. Custom functions go beyond CSS variables by accepting multiple arguments, supporting default values, and encapsulating reusable logic that returns a single output. Together, these additions aim to reduce redundancy in stylesheets and make conditional styling more explicit and maintainable. The if() function is currently available in Chrome starting from version 137.

0 comments Read more at DEV Community