Israeli Team Develops AI-Powered Browser, Ace, with Millions of Downloads
Translated & summarized from NEWSru Israel by baba
Israeli ad-tech firm Arpeely has launched the AI-powered browser Ace, developed by a team of ten, which has garnered 3.8 million downloads and 550,000 monthly active users. The browser, built on Chromium, utilizes AI agents to automate code updates and feature development, significantly reducing development time. Arpeely aims for this AI-driven approach to allow its small team to compete with global browser giants.
The story in 5 lines · by baba
- Israeli company Arpeely launched the AI-powered browser Ace with 3.8 million downloads.
- A small team of ten developed the browser, featuring an integrated AI assistant.
- AI agents automate code updates and bug fixes, reducing development time significantly.
- The browser is built on the open-source Chromium project, facing integration challenges.
- Arpeely aims to compete with major browser developers through its AI-driven process.
Israeli ad-tech company Arpeely has developed a new browser named Ace, featuring an integrated artificial intelligence assistant. The project was undertaken by a small team of just ten individuals. According to company data, Ace has been downloaded 3.8 million times, with a monthly active user base reaching 550,000.
Roi Shafer, Arpeely's Vice President of Research and Development, revealed these details at the Evolve to AI 2026 conference in Tel Aviv. He explained that about 18 months prior, the company, primarily focused on automated online advertising procurement systems, decided to create its own browser and AI assistant. The goal was to capitalize on evolving internet user habits and offer an alternative to Google Chrome.
Ace is built upon the open-source Chromium project, the same foundation used by Chrome and other browsers. A significant challenge involved consistently integrating Chromium updates, including security patches. One update required the team to manage approximately 74,000 changes and 169 code conflicts.
To overcome this, Arpeely developed an AI agent that automatically detects conflicts, tests browser functionality, and attempts to fix errors, escalating only the most complex issues to human developers. This AI integration transformed the code base update process from a two-week task into a daily automated routine.
Additional AI agents are employed for developing new features. They analyze tasks from Slack, identify necessary code modifications, write and test the code, including running the browser on virtual machines. However, this automation led to a new bottleneck: programmers struggled to review AI-generated code. Arpeely responded by creating a system that learns from developer feedback to improve the AI agents' performance. This reduced the median time for changes to the main program version from 20 hours to four hours.
Future plans include developing infrastructure to update individual browser functions without releasing a new program version, enabling rapid testing of new ideas and interface variations. The built-in AI assistant is also being refined through automated analysis of user interactions, studying anonymized dialogues to identify issues and needed features. Shafer emphasized that AI alone doesn't guarantee faster development; a holistic approach to software creation, addressing code writing, review, testing, and deployment, is crucial for small teams to compete with major browser developers.