Reliable knowledge stays traceable and open to correction.
Propperly maintains the current state of what a team knows — the decisions, definitions, evidence and open questions — as a human-reviewed, traceable current state that people and AI agents can build on instead of rebuilding each time.
What Propperly believes
Reliable knowledge is not a pile of everything ever said. It is a current state a team can stand behind — and it stays reliable only while it remains traceable to its evidence and open to correction when that evidence changes. A claim you cannot check, or cannot revise, is not something you can safely act on.
Most of what a team knows works against that. It scatters across people, code, dashboards, documents, and AI sessions: decisions get reversed three messages deep in a thread, a definition changes in one place and stays old in three others, and a path the team ruled out resurfaces months later as a fresh idea. No single place holds what is currently true — and the faster AI helps teams produce work, the faster that real state scatters.
What Propperly is building
Propperly maintains an evidence-backed, human-reviewed current state of a project: the decisions and the evidence under them, what changed, what was ruled out, and what is still open. It reads selected analytical work already in place — analyses, definitions, dashboards, code, AI sessions — reconstructs the current understanding, and then a person confirms, corrects, or rejects it. Nothing becomes authoritative because an AI produced it.
The result is not a cleverer search over everything said. It is a maintained understanding that stays current as the work moves, keeps its own history, can be traced back to its sources, and changes only when a person approves the change — so returning to a project, or handing it to an agent, starts from what is true now.
Why Propperly starts with analytics
Organizational understanding has to begin somewhere, and analytics is the sharpest place to start. Its knowledge is made of unusually crisp atoms — metric definitions, decisions, and the evidence behind them — so what is currently true can actually be recorded. The cost of getting it wrong is immediate and legible: one metric defined two ways is two different numbers in two meetings. And it is exactly where teams are now pointing AI agents, at the very definitions those agents are quietly guessing. Starting there gives a first slice that is small, high-value, and measurable, with a governed core to widen outward from later.
Knowledge advances through criticism and correction
Propperly is built on an idea associated with the philosopher Karl Popper: that knowledge advances not by declaring things certain, but by stating them clearly enough to be challenged, and correcting them when the evidence changes. That principle is why explicit conflict, criticism, and revision are first-class in the product rather than problems to hide.
This is an inspiration, not an affiliation or a claim about the company's name: Propperly has no association with Karl Popper or any institution.
The direction shows up in how the product behaves, not only in what it says:
- Knowledge stays provisional. A maintained state is the current best answer, not a final one.
- Claims stay challengeable. Every entry keeps its provenance, so a statement can be checked against the evidence under it.
- Contradictions are information. When two sources disagree, that conflict is kept as an open question rather than quietly resolved.
- Progress happens by correction. Revisions supersede earlier claims on the record, preserving what changed.
- Human judgment stays necessary. Nothing becomes authoritative because an AI produced it; a person confirms, corrects, or rejects it.
Propperly is not designed to create an unquestionable source of truth. It is designed to maintain a traceable, challengeable understanding that improves when the evidence changes.
Propperly was founded by Assaf Moyal, following a decade working in and leading Analytics teams.
Join the Founding Beta, or start an Analytics pilot.
If you run a long-lived project through Claude Code, the Founding Beta is being prepared for you. If you lead an analytics team, we can talk about a focused Analytics pilot on your own definitions and decisions.