Explore the future of proactive AI in Frontend development, building web extensions that continuously analyze user input and deliver real-time feedback straight into the browser.
AM
by Andreea Miclaus
What does the title even mean?
And what is this talk about?
Table of Contents
What are browser extensions, their popularity and how they work
A tech-y breakdown of the process behind
The new use case of AI in browser extensions
💀TLDR
Browser extensions are cool again & now with USEFUL GenAI features.
Chatbots are evolving from reactive tools to proactive assistants living inside your browser.
I built something similar to Grammarly but with AI in 3 months and now it makes sense why they are a full team doing it for longer (:
Browser extensions are actually popular?
What companies say
93%
Enterprise Adoption
Of enterprise companies utilize browser extensions daily.
70%
AI-Powered
Highlight AI extensions as the most popular category.
83%
Growing Usage
Of companies plan to increase their use of extensions.
What users say
30%
User Adoption
Of internet users install at least one browser add-on.
70%
Browser UX
Of users report that extensions enhance their browsing experience.
76%
Productivity Boost
Of users notice an increase in productivity when using extensions.
Privacy concerns
Anatomy of a Web Extension
manifest.json
The brain of your extension
UI Components
Popups, options pages, sidebars
Content Scripts
JavaScript injected into web pages
Background Scripts
Run in the background, listening for events
Manifest V3
Most extensions now use MV3 with background scripts as service workers, better performance and stricter security.
Content Scripts
Injected into web pages like dropping your own JS into someone else's site.
Read/modify the DOM
Sandboxed from page JS
Great for overlays and tooltips
Background Scripts
Run in the background, listening for events outside the DOM.
Coordinate data
Manage storage
Handle browser events
Popup
User interface component.
The little window that opens when a user clicks your extension icon.
Built with regular HTML/CSS/JS.
Acts like a tiny SPA — React or any frontend lib works here.
Messaging System
Extensions need communication. Either:
Popup ⇄ Background
Content Script ⇄ Background
Popup ⇄ Content Script (via background)
For this we use chrome.runtime.sendMessage() or ports
Permissions
Extensions need explicit permissions for everything — tabs, storage, cookies, etc.
These are declared in manifest.json
Other assets
Images for icons, fonts, and any bundled assets go here.
Declare what you need in the manifest.json file.
There is some more stuff…
(…but I drew this nice diagram and wanted a slide for it)
This was a proactive AI that did not wait around to be asked anything
Conventional (Reactive) vs. Proactive AI Approaches
1
"Explain"
You're actively prompting the AI
2
"Explain this"
You're giving context alongside the prompt
3
"Watch this"
Your actions are the prompt
From Searching for Answers to Answers Finding You
Picture joining a year-long project mid-stream. The knowledge is scattered across Notion, Slack, Google Docs, and countless tools.
What you would do
Ping teammates constantly with questions.
Go on endless scavenger hunts across platforms.
Manually connect disjointed information.
What proactive AI does
Tools like Glean work quietly as browser extensions.
They understand your company's internal knowledge.
Information finds you before you even ask.
Glean connects the dots as you work.
It offers relevant links, summaries, and connects you with people who've worked on similar things—all without explicit prompting.
This shift represents the future: answers searching for you, not you searching for answers.
Liner
Evolved from a web highlighter to an everything research tool with instant summaries and question answering.
JasperAI
Brings powerful content generation directly into Gmail, Notion, and LinkedIn with real-time assistance.
Glasp
Turns AI into a personalized learning engine that highlights, summarizes, and expands content.