ValidAnytime
  • Try it free
  • Observatory
  • Pricing
  • Docs
Sign inStart free
ValidAnytime

A guaranteed false-alarm budget across your whole fleet, valid no matter how often you look.

Made by Compiled Intelligence — a frontier AI lab working on quantitative finance from first principles; ValidAnytime is the monitoring we built for our own model fleets, productized.

Product

  • How it works
  • Pricing
  • Observatory
  • Try it free
  • Dashboard

Solutions

  • Who it's for
  • For LLM engineers
  • For MLOps engineers
  • For SRE & platform
  • LLM eval regression
  • Model drift detection
  • Alert fatigue

Developers

  • Docs
  • Quickstart
  • Python SDK
  • API reference

Learn

  • Benchmark
  • Detector guides
  • Glossary
  • Compare
  • Why dashboards lie
  • Blog
  • Anytime-valid 101
  • Winter Storm Uri case study
  • Incident ledger
  • Observatory methodology

Company

  • Contact
  • Privacy
  • Terms

Observatory alarms & product updates, by email

Rare emails only. We store your address and the tag “footer” — nothing else. Unsubscribe anytime.

© 2026 ValidAnytime. All rights reserved.

Every alarm ships with its guarantee_tag and theorem_ref.

Glossary

Anytime-valid inference

Anytime-valid inference is a way of testing that stays statistically valid no matter how often you look at the results.

Also known as: always-valid inference

Anytime-valid inference is a way of testing whose guarantees hold at every look simultaneously, so you can peek as much as you want and still trust the answer. If you watch a metric and check it whenever you feel like it — every minute, every deploy, every time an alert nags you — ordinary statistics quietly break. Each extra look is another chance to be fooled by noise, so your false-alarm rate creeps up without you noticing — anytime-valid inference is the fix.

For someone on-call, this is the difference between an alert you can act on and one you learn to ignore. When an anytime-valid monitor stays quiet, that silence is meaningful; when it fires, the error guarantee is exactly what it was designed to be — not something you eroded by refreshing the dashboard.

Under the hood it is powered by e-processes and confidence sequences, whose validity comes from Ville's inequality rather than from a fixed sample size. That is what makes "valid no matter how often you look" a theorem instead of a hope.

Go deeper

  • Anytime-valid 101 in the docs
  • Why your dashboard is lying to you
  • The honest drift-detector benchmark

Related terms

  • E-processAn e-process is a running score of evidence against 'nothing has changed'; its value at any moment is an e-value, and it stays valid at every look.
  • Confidence sequenceA confidence sequence is a sequence of confidence intervals that stays valid at every point in time, so you can read it whenever you like.
  • The peeking problemThe peeking problem is the reason a metric can look 'significant' just because you checked it too many times.
  • Sequential testingSequential testing is the practice of testing a hypothesis as data arrives, deciding to stop as soon as the evidence is conclusive.
  • Online FDR controlOnline FDR control is a way to bound the fraction of false alarms across many streams that you are testing continuously over time.
  • Conformal monitoringConformal monitoring is the practice of turning a model's outputs into calibrated evidence of change without assuming how the data is distributed.
  • Always-valid p-valueAn always-valid p-value is a p-value you are allowed to read at any moment — it never gets less trustworthy the more often you check.
  • Ville's inequalityVille's inequality is the theorem that makes 'valid no matter how often you look' true rather than wishful.
  • Test martingaleA test martingale is a running evidence score that, if nothing has changed, is not expected to grow — the honest core of an e-process.
  • False discovery rateThe false discovery rate is the fraction of fired alarms that turn out to be false — the number you actually want to control across a fleet.
  • Changepoint detectionChangepoint detection is the task of spotting the moment a metric's behavior genuinely shifts — as opposed to normal noise wobbling around.
  • SPC for MLSPC for ML is the practice of putting machine-learning metrics under statistical process control — control charts with explicit rules, not eyeballed dashboards.

Put the theory to work.

ValidAnytime turns these ideas into a live alarm you can trust — valid no matter how often you look. Prove it on your own data, free.

Start freeTry the detector