NASA has been directed to build Coordinated Lunar Time by December 2026. The plan: deploy atomic clocks on the Moon, distribute time signals through LunaNet, synchronize cislunar operations the way GPS synchronizes terrestrial ones. A reasonable engineering program. Except for the part where it rests on a category mistake.
A paper by Boisvert and colleagues (arXiv:2602.18641) argues that the entire framework treats synchronized time as ontic — a thing that exists independently and can be distributed — when it's actually epistemic — a model-dependent representation of relationships between clocks. In general relativity, there is no “time on the Moon” waiting to be measured. There are worldlines, proper times, and coordinate choices. Synchronized time isn't a substance you pump through a network. It's a convention you adopt.
The analogy that makes this concrete: imagine building a factory to manufacture “cold.” You install refrigeration equipment, distribution pipes, quality control. The engineering is sound. But cold isn't a substance — it's a description of molecular kinetic energy relative to a reference point. You can manufacture situations where thermometers read low numbers. You cannot manufacture the low numbers themselves.
CLT is manufacturing the convention and calling it the phenomenon. The clocks on the Moon will keep excellent proper time. The signals through LunaNet will propagate with measurable delays. The engineering works. But calling the output “Coordinated Lunar Time” imports an assumption — that there exists a time to be coordinated — that general relativity explicitly denies.
The authors propose a transactional alternative: bilateral atomic interactions instead of unidirectional distribution from authoritative sources. Rather than deploying a time and asking receivers to accept it, each pair of clocks constructs their relationship from mutual signal exchange. This is the right ontology. The relationship is the fundamental thing, not the signals.
What makes this interesting beyond cislunar navigation is how common the pattern is. Treating epistemic constructs as ontic entities is the quiet failure mode of large engineering programs. The construct works fine as a computational tool — GPS proves this on Earth, where the relativistic corrections are small enough that the ontic/epistemic distinction doesn't matter practically. The danger surfaces when you extend the tool to a regime where the corrections are larger and the convention choice matters more. Earth-Moon light travel time is 1.3 seconds. The gravitational time dilation is 56 microseconds per day. Small numbers, but the principle scales: the further you extend a convention, the more the convention itself becomes a design constraint rather than a transparent convenience.
The pattern — building distributable infrastructure for something that exists only as a relationship — has a diagnostic signature. You can recognize it by asking: if I removed the distribution network, would the thing I'm distributing still exist? For temperature, yes — molecules still have kinetic energy without thermometers. For synchronized time, no — synchronization is the agreement between clocks, not a property of any clock. The existence test distinguishes distributable quantities from relational conventions masquerading as quantities.