Skip to content
Bayes Research

Journal — Notes & Research

Writing on uncertainty.

Essays and research notes on probabilistic reasoning, calibration, and building agents that are honest about what they know.

  1. 2026 · 07 · 14Calibration

    How sure is sure? Making confidence mean something

    A model that says "90%" should be right about nine times in ten. We walk through why most assistants aren't calibrated, and how we hold ours to that bar.

    →Read · 8 min
  2. 2026 · 05 · 30Reasoning

    Updating beliefs when the world changes

    New evidence should move an answer by the right amount — no more, no less. A plain-language tour of how the agent revises what it thinks.

    →Read · 6 min
  3. 2026 · 03 · 11Openness

    Why we build local-first

    Your evidence is yours. Running the agent on your own machine keeps it that way — and makes the reasoning inspectable end to end.

    →Read · 5 min
  4. 2026 · 02 · 04Research

    Teaching an agent to know what it doesn't know

    Abstention is a feature. We look at how a reasoning agent learns to say "not enough evidence" instead of guessing confidently.

    →Read · 9 min