A project of Old Watchman
The world before Abraham,
through text, archaeology, and the earth itself.
An interactive research platform that holds sources rather than conclusions.
Where the evidence disagrees, the disagreement is preserved. Where we hold nothing,
the page says so by name.
Evidence before belief
This is not an argument. It is a library with a memory of where every statement came from.
Each claim in it can be walked back: to the witness it rests on, the line of that witness,
the edition behind the line, and the rights under which we hold that edition.
When a chain breaks, the break is displayed rather than hidden.
What we currently hold
Counts are read live from the database on every page load. Nothing here is
a marketing figure; a zero is displayed as a zero.
The one rule everything else follows
The database was not designed to prove a theory.
It was built to hold evidence in categories that cannot be
quietly merged, so that a reader can see what kind of thing each statement is:
- Never erase disagreement. Where sources conflict, both remain, labelled.
- Never silently merge identities. Two names that might be the same person are two records until a human says otherwise.
- Never turn tradition into archaeology. A text naming a city does not excavate it.
- Never turn correlation into proof. Two things dated alike are two things dated alike.
- Never turn a period label into a measurement. "Uruk period" is a classification, not a date.
Where to go
What this platform refuses to do
Unknown location
Does not get an invented pin.
Unknown date
Does not get an invented point on the timeline.
Missing manuscript
Is not called "absence" — it is called not held.
Restored text
Is not presented as preserved text.
Translation wording
Is not presented as the ancient word.
Unheld publication
Is not displayed as held evidence.
A recent example, recorded in full
A date of "c. 2900 BC" for a flood deposit at Shuruppak appeared in this
project's own early design material. When the archaeological evidence chain was finally
built and run, it returned zero radiocarbon samples, zero stratigraphic
contexts and zero datable events for that site — and no source anywhere in the database
supported the number.
The number had been written from general knowledge during interface
design. It is now recorded as UNSOURCED — not
refuted, and not confirmed. The correction is kept visible on the
About page rather than quietly deleted, because the mistake is part of
the record too.