I built an AI QA teammate. Here's what it does.
Bytesalt is in Research Preview. This is what it is and why I built it.
Today I’m putting Bytesalt into Research Preview.
This post is a quick explanation of what it is and why I built it.
What the problem is
I spent years as an engineer at AWS and PayPal — environments where the cost of a production bug you didn’t anticipate is very real. That’s where I started noticing how much slips through even well-maintained test suites.
The pattern is always the same: your suite passes, CI is green, and something still breaks for real users.
Not because your Playwright or Selenium tests are bad. They’re not. They’re doing exactly what they’re supposed to do — validating that the paths you thought of still work the way you told them to.
The problem is everything else.
The support widget that blocks the ‘Place Order’ button on mobile. The cookie edge case that puts users in a weird state. An authenticated API endpoint still returns data after the session token is expired. The button that looks fine in light mode — but disappears in dark mode.
These bugs are common, they’re real, and they span a much wider surface than functional correctness — security, accessibility, cross-browser behaviour, mobile, UX. They’re genuinely hard to cover with scripted tests, not because scripted tests are primitive, but because you can’t write a script for something you haven’t already imagined.
What Bytesalt actually does
You give it a plain English goal: "Test the checkout flow". Bytesalt distributes that across parallel AI agents that actually browse your application the way a real user would. They explore, they click things, they try weird combinations. And they’re not all looking for the same thing — one is checking accessibility, another is probing for security issues, another is testing across browsers and devices, another is evaluating UX friction. When something surfaces, you get a report that tells you what broke, why it broke, and how to fix it.
It’s not a replacement for Playwright or similar tools. Your deterministic tests still run. Bytesalt is the layer that handles the unknown — the workflows you didn’t think to write scripts for.
A few practical things worth knowing:
- You can run it from the CLI, or plug it into your CI/CD pipeline alongside your existing tests.
- If your app is on localhost or a private network, there’s tunnel support built in. No firewall changes needed.
- Jobs run in parallel, so even deep tests don’t take hours.
How to try it
Install the CLI and run your first test with free credits:
macOS / Linux / WSL
curl -fsSL https://bytesalt.com/install.sh | shWindows (PowerShell)
irm https://bytesalt.com/install.ps1 | iexThen:
bytesalt start "Test https://www.craigslist.org for usability on an iPhone. Check only above the fold. Report a single issue and stop. Do not explore."The Quick Start guide has more details, including how to test private apps and how to set up CI/CD integration.
No credit card required to try it.
If you run it and have thoughts, I’d love to hear them — deepank@bytesalt.com or via the contact page.