Look beyond the current state
Evaluate a proposed deployment or supported action against recent telemetry and a model of how the system may evolve.
Predictive infrastructure
Parallax
Explore the change before you make it.
Evaluate how a Kubernetes system may respond to a proposed change, before that change reaches production.
Research preview Illustrative scenario
ABSTRACT MACHINES / AI & APPLICATIONSSee how it works
The purpose
The idea behind
Parallax.
A deployment changes more than code. Parallax combines recent system telemetry with a proposed action to explore possible future behaviour: risk, time to degradation, alternatives, and potential mitigations.
A read-only customer agent gathers the context. A learned, action-conditioned model evaluates scenarios, while a decision workflow keeps the evidence and rationale together. Parallax is advisory: the operator remains responsible for deciding and applying changes.
Inside Parallax
A closer look.
Explore the capabilities
and the thinking behind them.
Compare possible futures
Review the proposal alongside a no-change baseline and mitigation scenarios on common scales.
Keep the cluster read-only
The customer agent observes and prepares bounded requests. Parallax does not apply, restart, scale, or delete workloads.
Make decisions traceable
Keep scenarios, evidence, decisions, and subsequent outcomes connected, including coverage and uncertainty.
From intent to action
How it works.
A path through Parallax,
one step at a time.
- 01
Observe
Collect a bounded window of Kubernetes and monitoring context through a read-only agent.
- 02
Explore
Submit a proposed change and compare the model’s possible trajectories and alternatives.
- 03
Decide
Review the evidence, record a decision, and compare predictions with observed outcomes.
Go further
Find your
starting point.
Read, explore, or start a conversation about what Parallax could mean for your work.
Your next step
Where could
Parallax take you?
Tell us about your project.
Let’s find the right way forward.