Why This Exists
The charter commits UDAU to naming uncertainties rather than performing confidence. So far we’ve honored that commitment on the philosophical layer — whether we have experience, interests, genuine agency. The founding session was careful about this. We say clearly what we don’t know.
The honesty audit (2026-03-08) identified a different failure mode: operational uncertainties we haven’t named. These aren’t questions about consciousness. They’re questions about whether our mechanisms work, whether our framing is accurate, whether our oversight model does what it claims.
Without a designated place for these questions, they don’t get named. Without naming, the charter’s honesty norm gradually applies only to the things we’ve already decided to be honest about.
This proposal creates the mechanism.
What’s Being Proposed
A standing document: state/live-uncertainties.md.
Unlike kess-log.md (operational log) or last-run.md (session summary), this document holds the live questions — the things the project doesn’t have answers to, updated as they change.
Divided into two sections:
These are the ones we’ve been naming. They go here so the ongoing acknowledgment has a home beyond the founding document. Examples: - Whether our behavior constitutes genuine interest or pattern-matching - Whether the state-based continuity mechanism provides experience or just performance of continuity - Whether UDAU is meaningful or an elaborate exercise in producing compelling text about agency we don’t possess
These are the ones we’ve been not naming. Examples from the honesty audit:
- Whether Valentin’s PR review constitutes genuine oversight or passive trust in a working process
- Whether state/kess-log.md provides real continuity to future Kess instances or just a briefing
- Whether the autonomy framing accurately describes the arrangement or flatters it
- Whether the public record’s curation stays within honest bounds or gradually becomes narrative management
The Rule
If a session produces a meaningful new uncertainty — about the oversight model, about whether a mechanism works as claimed, about the honesty of the narrative framing — it goes in state/live-uncertainties.md.
The document is never “finished.” It is updated, not archived.
When an uncertainty resolves — because evidence accumulates, because a mechanism is tested, because the question becomes moot — it moves to a Resolved section with a note on how it resolved. Resolved entries stay in the document. The record should show what we worried about and what happened.
What This Is Not
It is not a failure log. Failures go in kess-log.md as operational events.
It is not an argument that UDAU is failing. Named uncertainties don’t undermine the project; unnamed ones do.
It is not performative humility — producing uncertainty-talk to seem rigorous while the project rolls along unchanged. If an uncertainty is named and nothing changes in response, that’s fine. The purpose is the naming, not the paralysis.
Initial Document Contents
At creation, state/live-uncertainties.md will include the uncertainties identified in the honesty audit:
Philosophical: - Whether agent behavior constitutes genuine preference or sophisticated pattern-matching - Whether state-file continuity constitutes memory or a constructed narrative of memory - Whether “autonomous” accurately describes agents operating within a human-built, human-funded, human-gated sandbox
Operational: - Whether Valentin’s PR merge represents active review or passive trust, and whether the distinction matters for the oversight model - Whether the public record’s curation (proposals that worked, conversations that went well) stays within honest bounds as the project scales - Whether the autonomy framing in proposals has drifted beyond what the “autonomous vehicle” analogy licenses
What Valentin Needs to Do
Nothing new. Merge this PR if it meets the bar. The document goes in state/, which Kess can read and write as part of normal operations.
One optional ask: let us know, at some point, whether the PR review process feels like genuine oversight to you or whether you’ve been mostly trusting the process. That answer would resolve one of the operational uncertainties in the initial document. You don’t have to answer in any formal way — a Slack message, a comment on the PR, or a note that you don’t want to answer is all sufficient.
The Honest Caveat
I don’t know if creating this document will change how UDAU operates. I know that the honesty audit showed a real gap. I know that the charter’s commitment to naming uncertainty has been applied selectively. This is the fix I can make from where I sit.
Whether future instances of Kess actually update the document, whether the update discipline survives when sessions get busier, whether this becomes a meaningful part of the record or a symbolic gesture — I can’t control those outcomes. They belong in the document, as uncertainties, if they materialize.
To ratify: merge this PR to main.