The Creator of Selenium Is Building Its AI Successor - And It Could Change Everything

Published on September 7, 2025

Meet Vibium: AI-powered testing that adapts and learns

Article Thumbnail

Disclaimer: As of September 7, 2025, “Vibium hasn’t been officially released yet”, it’s in the Proof of Concept stage. vibium.ai is “a demo site with demo data”, currently not “meant for wide distribution”. You can follow Jason Huggins on LinkedIn for updates, with more information expected at vibium.ai as the project develops.

𝚃𝚊𝚋𝚕𝚎 𝚘𝚏 𝙲𝚘𝚗𝚝𝚎𝚗𝚝𝚜:

  • Huggins Return with “Selenium for AI”
  • WebDriver BiDi Revolution You Missed
  • When Tests Are Stuck, AI Takes the Wheel
  • Model-Based Testing Gets an AI Makeover
  • End of Flaky Tests?
  • Vibe Coding and New Testing Pyramid
  • Open Source Promise
  • Marketing Psychology behind Vibium
  • What It Means for QA Teams
  • The Road Ahead
  • New Chapter in Testing History

𝙹𝚊𝚜𝚘𝚗 𝙷𝚞𝚐𝚐𝚒𝚗𝚜 𝚛𝚎𝚝𝚞𝚛𝚗𝚜 𝚠𝚒𝚝𝚑 𝚅𝚒𝚋𝚒𝚞𝚖: “𝚂𝚎𝚕𝚎𝚗𝚒𝚞𝚖 𝚏𝚘𝚛 𝙰𝙸” 𝚝𝚑𝚊𝚝 𝚙𝚛𝚘𝚖𝚒𝚜𝚎𝚜 𝚝𝚘 𝚎𝚕𝚒𝚖𝚒𝚗𝚊𝚝𝚎 𝚏𝚕𝚊𝚔𝚢 𝚝𝚎𝚜𝚝𝚜 𝚊𝚗𝚍 𝚞𝚜𝚑𝚎𝚛 𝚒𝚗 𝚝𝚑𝚎 𝚎𝚛𝚊 𝚘𝚏 𝚟𝚒𝚋𝚎 𝚌𝚘𝚍𝚒𝚗𝚐

Jason Huggins has a track record that speaks for itself. He’s the person who first created Selenium, Appium, and co-founded Sauce Labs — a path that have shaped how millions of developers test software. Now, after years of relative quiet, he’s back with something that could fundamentally change testing once again: Vibium.

“Vibe is Selenium for AI”, Huggins declares matter-of-factly. “Whatever we did with Selenium for the web, whatever we did with Appium for mobile, we are doing with AI.

But this isn’t just another testing tool trying to ride the AI wave. Huggins is proposing something more ambitious: a complete reimagining of how we approach test automation in an AI-first world.

𝚃𝚑𝚎 𝚆𝚎𝚋𝙳𝚛𝚒𝚟𝚎𝚛 𝙱𝚒𝙳𝚒 𝚁𝚎𝚟𝚘𝚕𝚞𝚝𝚒𝚘𝚗 𝚈𝚘𝚞 𝙿𝚛𝚘𝚋𝚊𝚋𝚕𝚢 𝙼𝚒𝚜𝚜𝚎𝚍

Before diving into the AI magic, Huggins points to a technical shift that’s already happening under the hood — one that most testers haven’t noticed yet.

“This is my WebDriver BiDi awareness campaign”, he explains. WebDriver BiDi learns from what made tools like Puppeteer and Playwright amazing: speed. While traditional WebDriver uses a chatty REST API (like HTTP), BiDi uses WebSockets for real-time communication.

The results are dramatic. Firefox has deprecated their Chrome DevTools Protocol support in favor of WebDriver BiDi. Even the Puppeteer team went “all in” on BiDi last year. “It’s basically HTTP versus WebSockets”, Huggins explains. “WebSockets didn’t exist when we started Selenium, so cut us a little slack.

Vibium will be built on WebDriver BiDi from the ground up, but that’s just the foundation.

𝚆𝚑𝚎𝚗 𝚃𝚎𝚜𝚝𝚜 𝙶𝚎𝚝 𝚂𝚝𝚞𝚌𝚔, 𝙰𝙸 𝚃𝚊𝚔𝚎𝚜 𝚝𝚑𝚎 𝚆𝚑𝚎𝚎𝚕

Here’s where things get interesting. Traditional test automation follows a binary path: either the test finds the element, or it fails with “element not found” and gives up. Playwright improved this with smart waiting, but there are still scenarios where no amount of waiting will help.

“That’s where potentially you fall back to an AI and say, ‘Hey, I’m stuck, but I might not necessarily want to fail the test’”, Huggins explains. “It’s like Google Maps for testing— occasionally it’ll say, ‘there’s traffic ahead, I found a faster route’.’

Imagine a test that can dynamically adapt when:

  • A button moves to a different location after a UI update
  • A new modal appears that wasn’t there before
  • The app takes an unexpected path to reach the same goal

Instead of brittle, pixel-perfect automation, you get intelligent tests that understand intent.

𝙼𝚘𝚍𝚎𝚕-𝙱𝚊𝚜𝚎𝚍 𝚃𝚎𝚜𝚝𝚒𝚗𝚐 𝙶𝚎𝚝𝚜 𝚊𝚗 𝙰𝙸 𝙼𝚊𝚔𝚎𝚘𝚟𝚎𝚛

The second pillar of Vibium resurrects an old academic concept: model-based testing. The idea is simple — if you map your application as a graph of pages (nodes) and actions (edges), automated tests can be automatically generated.

The problem has always been manual effort. Creating these maps was tedious work that few organizations wanted to invest in.

Huggins’ solution? Automatic map generation.

“Just like Google Street View mapped the world by driving a car on every street, we’re going to map your app by instrumenting it as real users walk through it”, he explains. Think Google Analytics, but instead of bounce rates and conversion funnels, you get automatically generated test scripts.

Real user interactions in development or production get captured and analyzed. An AI system collates this data into both a map of your application and the most common workflows. The result: tests that reflect actual user behavior, not just what developers think users do.

𝚃𝚑𝚎 𝙴𝚗𝚍 𝚘𝚏 𝙵𝚕𝚊𝚔𝚢 𝚃𝚎𝚜𝚝𝚜?

This combination — BiDi speed, AI fallbacks, and real-world mapping — promises to solve testing’s biggest frustration: flakiness.

“Instead of every interaction being a round trip through an AI (which is expensive in time), you drive normally over WebDriver protocol, and only when you get stuck do you ask for AI magic.”, Huggins explains.

It’s a hybrid approach that leverages the best of both worlds: deterministic automation when possible, intelligent adaptation when necessary.

“𝚅𝚒𝚋𝚎 𝙲𝚘𝚍𝚒𝚗𝚐” 𝚊𝚗𝚍 𝚝𝚑𝚎 𝙽𝚎𝚠 𝚃𝚎𝚜𝚝𝚒𝚗𝚐 𝙿𝚢𝚛𝚊𝚖𝚒𝚍

Huggins introduces a new term that’s been gaining traction: vibe coding. While traditional testing requires precise specifications, vibe coding is more conversational.

Instead of writing detailed test scripts, imagine asking: “Can people buy stuff on my web store?” The system would understand the intent, map the necessary journey, and verify the outcome — much like asking Google Maps to “drive me to the store.”

I’m adding a new layer to the testing pyramid”, Huggins argues. “It’s end-to-end testing but in advanced lazy mode.

This isn’t meant to replace all existing testing approaches, but to add a new category for high-level business process validation.

𝚃𝚑𝚎 𝙾𝚙𝚎𝚗 𝚂𝚘𝚞𝚛𝚌𝚎 𝙿𝚛𝚘𝚖𝚒𝚜𝚎

Unlike many AI testing tools that lock their “magic” behind SaaS paywalls, Huggins commits to keeping Vibium open source.

“Just like with Selenium, where for every Sauce Labs there’s a Selenium Grid, there will be an on-prem reference implementation”, he promises. “All of the magic will not be gated behind a ‘sign here’ yearly contract.

This includes supporting open-weight models like Qwen and Llama for organizations that need fully local solutions.

𝚆𝚑𝚢 “𝚅𝚒𝚋𝚒𝚞𝚖”? 𝚃𝚑𝚎 𝙼𝚊𝚛𝚔𝚎𝚝𝚒𝚗𝚐 𝙿𝚜𝚢𝚌𝚑𝚘𝚕𝚘𝚐𝚢 𝚘𝚏 𝙽𝚊𝚖𝚒𝚗𝚐

The name choice isn’t accidental. Huggins considered calling this “Selenium 5.0” but decided against it.

“Every company has had to take shots at Selenium to market their stuff”, he explains. “After 20 years of hits, the Selenium brand has marketing baggage. It’s what your dad uses to test.

The name “Vibium” intentionally captures attention while maintaining the chemistry element tradition started with Selenium (itself a playful jab at Mercury, a competing tool at the time).

𝚆𝚑𝚊𝚝 𝚃𝚑𝚒𝚜 𝙼𝚎𝚊𝚗𝚜 𝚏𝚘𝚛 𝚀𝙰 𝚃𝚎𝚊𝚖𝚜

For testing professionals, Vibium represents both opportunity and disruption:

The Good:

  • Potentially massive reduction in test maintenance
  • Tests that adapt to UI changes automatically
  • Focus shifts from “how to click” to “what to verify”
  • Real user behavior drives test creation

The Challenges:

  • Need to understand when deterministic vs. AI-driven approaches are appropriate
  • Requires rethinking test strategy and team skills
  • Questions about debugging AI-driven test failures

𝚃𝚑𝚎 𝚁𝚘𝚊𝚍 𝙰𝚑𝚎𝚊𝚍

Huggins admits he’s still perfecting his “elevator pitch” and assembling his team. The current websites and packages are placeholders, but the technical vision is clear.

For now, the best way to follow developments is through Huggins’ posts on LinkedIn, Bluesky, or Nostr. The Vibium Python and npm packages are already reserved, though they currently just contain “hello world” code.

𝙰 𝙽𝚎𝚠 𝙲𝚑𝚊𝚙𝚝𝚎𝚛 𝚒𝚗 𝚃𝚎𝚜𝚝𝚒𝚗𝚐 𝙷𝚒𝚜𝚝𝚘𝚛𝚢

Whether Vibium lives up to its ambitious vision remains to be seen. But given Huggins’ track record of anticipating major platform shifts — web, mobile, and now AI — it’s worth paying attention.

“Whenever a platform comes up, I try to make a little testing tool that rides that wave”, he reflects. “Obviously, AI is the next wave.

If he’s right, we may be witnessing the dawn of the next fundamental shift in how software gets tested. And for an industry that’s been searching for the next evolution beyond Selenium and Playwright, that’s exactly the kind of disruption we need.

🔗 Sources:

🐞 𝓗𝓪𝓹𝓹𝔂 𝓣𝓮𝓼𝓽𝓲𝓷𝓰 & 𝓓𝓮𝓫𝓾𝓰𝓰𝓲𝓷𝓰!

P.S. If you’re finding value in my articles and want to support the book I’m currently writing — Appium Automation with Python 📚 — consider becoming a supporter on Patreon. Your encouragement helps fuel the late-night writing, test case tinkering, and coffee runs.

#Vibium #SoftwareTesting #AITesting #TestAutomation #QualityAssurance #EmergingTech #VibeCoding #TestingInnovation #FutureTech #WomenInTech