The Crux Registry

The Crux Registry

Open a modern Bible to Acts 8 and read the story of the Ethiopian eunuch. Somewhere around the baptism, watch the verse numbers: 36, then 38. Verse 37 — the eunuch's confession, I believe that Jesus Christ is the Son of God — is simply not there. A footnote at the bottom of the page says something like "some manuscripts add," and the text moves on.

That footnote is doing an enormous amount of work. Behind its eight words stand actual manuscripts with actual dates — which ones have the verse, which ones don't, how old each is, and where a curious reader could go look. Almost nobody gets past the footnote, because the full answer has always lived somewhere else: in the apparatus at the bottom of scholarly editions, written in compressed Latin abbreviations, in books most Christians will never open and were never really meant to.

I've spent the last week building the place I always wanted that footnote to lead. It's called the Crux Registry, it covers sixty-five of the most famous disputed readings in the Bible, and it's free and public right now.

What a Crux Is

Scholars call these places cruces — from the Latin for cross, because a crux interpretum is where the interpreter gets hung up. A textual crux is a spot where the oldest manuscripts of the Bible genuinely disagree: over a letter, a word, or an entire passage.

Some of them you already know, even if you've never heard the term. Whether Mark's Gospel originally ended at 16:8, or included the twelve verses of snake-handling and ascension that follow. Whether the woman caught in adultery — arguably the most beloved story in the Gospels — was in John's Gospel to begin with. Whether the number of the beast is 666 or 616 — the earliest surviving witness to that verse is a third-century papyrus, and it reads 616.

Others are smaller and stranger. In Psalm 22:16, the difference between "they pierced my hands and feet" and "like a lion my hands and feet" is a single Hebrew letter — a vav versus a yod, the length of one downstroke. In Genesis 4:8, the Hebrew text says "Cain said to his brother Abel" — and never says what. In Deuteronomy 32:8, the nations are divided according to the number of the "sons of God" or the "sons of Israel," depending on whether you follow a Dead Sea Scroll or the medieval Hebrew tradition, and the two readings open onto very different pictures of the heavenly court.

Every Witness on the Record

Each crux gets one page, and every page is built the same way. The two readings sit side by side at the top, in Hebrew or Greek, with transliteration and a plain gloss. Below them is the witness table: every manuscript, ancient translation, and critical edition on the record for that reading — what it actually says, how old it is, and a link to the published source so you can check it yourself. Then the stakes, stated for both sides without a thumb on the scale. Then the related cruxes, because these things run in families.

The registry currently holds 176 witnesses across the sixty-five readings, organized into five sections: seventeen Hebrew readings, twenty-four Greek readings, fourteen whole verses that aren't in every Bible, four entire passages, and six numbers the manuscripts disagree about. If your Bible seems to skip a verse number, the reason is almost certainly on this list.

A word about how it was built, because the discipline matters more than the feature. Every witness row was curated by hand — none of the manuscript data is AI-generated. Where a source corpus exists on disk, a checker verifies the quoted reading letter for letter against it before the page is allowed to publish: Hebrew readings against the Westminster Leningrad Codex, Greek readings against the SBL apparatus. Witnesses with no corpus to check against — the Scrolls, the Septuagint, the ancient versions — carry their dates and links to the published editions instead. AI writes connective prose only, and a verifier audits that prose against the sources before anything ships. The facts on the page are never generated.

The Committee Didn't Know

There's one thing on these pages that I believe exists nowhere else.

The Anselm Project Bible was translated by a blind, role-separated committee — translating stages that were never told they were reading Scripture, an adversarial auditor back-translating everything they produced, and deterministic gates no model can talk past. That committee hit every one of these sixty-five forks in the road. It had no tradition to defend, no doctrine to protect, no idea that a downstroke in Psalm 22 had two thousand years of argument riding on it. It just had the tokens in front of it.

Each crux page shows what the committee did when it faced the same choice — the reading it followed, pulled live from the published release records, with the full audit trail public. I find these panels the most interesting thing in the registry, precisely because the committee's answer carries no authority at all. It isn't a church father or a text critic. It's something stranger: a witness with no stake in the outcome.

Zero Verdicts Is the Point

At the top of the registry there are three numbers: 65 readings, 176 witnesses, 0 verdicts. The zero is deliberate, and it's the part I care most about.

"The Bible has been changed" is one of the stock objections every pastor eventually hears. I wrote in the Paradigm Apologetics series that the stated objection is rarely the one doing the real work — but this particular objection deserves a real answer anyway, because the worst possible response is the nervous subject-change that confirms the suspicion behind it. The person who asks has usually just learned, from a YouTube video or a college course, that the manuscripts differ in thousands of places. Nobody told them before. The discovery feels like a cover-up unraveling.

But the church never hid this. The 1611 King James — the translation most often waved as the unchangeable original — printed a note in its own margin at Luke 17:36: "This 36. verse is wanting in most of the Greek copies." The translators knew. The differences have been on the record for centuries; the record was just kept in a form almost nobody could read.

So the registry doesn't argue. It lays the evidence out and lets you look, because this is exactly the kind of claim that does more work shown than argued. "Thousands of variants" sounds devastating as a headline. Then you walk through the worst cases one at a time — and they are all here, the most famous and most disputed readings in the whole Bible — and you see what they actually are: a downstroke, a missing doxology, a benediction that wanders between chapters, a confession some scribe loved too much to leave out. You will search the witness tables a long time for the crux on which the faith stands or falls. I don't think you'll find one — but I'd rather you weigh that against the evidence than take my word for it, which is why the verdict column stays empty.

In the Reader

The registry is also wired into the Bible reader. When you're reading a verse that carries a crux, the verse panel shows a card flagging it, with a line about what's disputed and a link to the full page. The footnote finally goes somewhere. The margins are now doors.

Go Look

The Crux Registry is live at anselm-project.com/bible/crux — free, public, no login. Start with Psalm 22:16 if you want to see how much can hang on one letter, or Revelation 13:18 if you want to surprise someone at Bible study. Every witness is on the record. The evidence is the point.

God bless, everyone.