Until now, when Antithesis found a bug, most customers waited until the end of the test run to start investigating.
That changes today.
When Antithesis finds a bug, we can now immediately notify you, or your agent, that a property has failed (or passed!), enabling you to start debugging failures right away.
You or your coding agent can use Multiverse Debugging to investigate in the live testing session, before the test finishes. Previously, a bug found fifteen minutes into an overnight run waited until morning to get your attention. Now, your agent can start investigating while Antithesis continues exploring your system and propose a fix before you wake up.
Getting started with this new workflow is simple: start a test, receive a property failure, and investigate it with Multiverse Debugging in the session that’s already running. Antithesis keeps on testing while you or your agent work to understand the failure. Once you’ve got a root cause and a fix, you can cancel the run, make a fix, and start a new test to confirm it.
This dramatically changes what an overnight test can now accomplish. Instead of waking up to failure in the morning and only starting the investigation then, you can have an agent working on it while you sleep. Multiple failures? No problem – multiple subagents can work in the same session. Failures found early in the run can get hours of attention before you return.
Until today, many users ran short, fifteen-minute tests to get quick feedback, then started longer runs overnight to explore more thoroughly. This change lets you act on early findings without shortening the run or waiting for it to finish. You no longer need to trade thoroughness for faster feedback, or use compute hours on short runs.
Longer runs still give Antithesis more time to explore your system. Streaming the properties to you in near real time makes the findings useful as they arrive: testing can continue, and debugging can begin.
Try it yourself #
Start an Antithesis test on release 66.
The results stream in the new run view on the web app, but this feature really shines if you’re working with agents.
Make sure you have the latest versions of the Antithesis skills and Snouty CLI, then prompt them to investigate and fix as well as running the test, e.g. “Run a test, investigate any bugs that surface, write a fix, and then validate them.”
Then go to bed!