Three Days and Three Nights

Working Out When Holy Week Happened, From the Text Itself

Being written.

About the Book

We call it Good Friday, and we celebrate the resurrection on Sunday. Jesus said he would be "three days and three nights in the heart of the earth." The math never worked for me. This is the most important week in Christianity. Why don't we just know when it happened?

So I went looking. It isn't that the Bible doesn't tell us. It isn't easy, but the days can be worked out from the text. The harder question was why it became a problem at all. That answer is written down too, three hundred years later, by an emperor and a council of bishops.

I wrote this because doubt about the timing turns into doubt about the event. And Holy Week happened.

Every verse and every source is cited, so you can check it yourself. Read it with your eyes wide open.

What's Inside

• The prophecies. What the Old Testament said before anyone knew how it would be used.
• The week. What happened, day by day, with every Gospel checked against the others, including the calendar math on the Thursday-or-Friday question.
• The politics. How the Church, centuries later, took a moveable Jewish festival and turned it into a fixed, required Sunday holiday, in the participants' own recorded words.
• Where the record runs out. What's left in the Easter you know today, and which layer each piece traces back to: the text, the calendar, or fourth-century politics.

The Tool

I spent about forty years building contextual search engines. For this book I built one more, pointed at scripture.

It's a search engine built on a library of primary sources. It runs on my own computer, and no outside service sees the searches or the library.

The Library

• 17 collections: the Hebrew Bible, Apocrypha, Pseudepigrapha, Dead Sea Scrolls, New Testament apocrypha, Josephus and the Jewish historians, ancient Near East texts, the Ethiopian Bible, Roman historians, the councils and early church writers, Gnostic texts, and classical Greek philosophy.
• 172 documents, split into 125,866 searchable passages. Scripture is split one verse at a time, so every result points to an exact verse.
• Every Hebrew and Greek word of the Old and New Testaments, all 31,161 verses, tied to its Strong's number and the original word: 682,470 tagged word parts in all. Pick a word and you see every other verse that uses it.
• 14,197 Strong's definitions and 3,961 Bible dictionary entries (Easton's, 1897), used as lookups only.
• Source texts beside the library: the Hebrew Old Testament (Westminster Leningrad Codex), the Greek New Testament, the Greek Old Testament (Swete's Septuagint), the Latin Vulgate, Dead Sea Scrolls Hebrew, and cuneiform texts.
• New sources don't go straight in. They're staged first, checked and graded, and only then added.

How a Search Works

• Exact words and meaning, together. One score finds the words you typed. A second finds passages that say the same thing in different words. The two are blended into one ranking.
• Related terms. A concept map of 70,729 links can widen a search with words you didn't type. The results show which terms matched, so you can see why a passage came up.
• Tags you can count. Passages are tagged by theme, by time period, and by when a writer says he's quoting ("as it is written," "spoken by the prophet"). Every tag is made by a rule applied to the text itself, so it can be checked.
• Two grades on every source. How much the research leans on it, and how well it holds up as a record. A source can matter a lot and still stand on thin ground, and the grades keep that visible.
• No outside commentary or study-Bible cross-references are built in. If a connection shows up, it's because the text makes it.

The AI Guardrails

I used AI to help build the tool and to work through what it turns up. AI is fast. Blazing fast. It isn't smart. The user has to be smart: know what they're looking for, and be able to pick out the results that answer the question.

The AI works under written rules:

• Sources only. Nothing outside the library feeds a finding.
• It can't decide. It can search, sort and draft. It can't decide what counts as a connection, and it can't state a conclusion as fact.
• Every claim carries its source. If there's no source, it asks instead of filling in the blank.
• Contradictions are speed bumps. When two accounts don't seem to agree, slow down, read it again and dig. Both verses go on the page.
• Check the Hebrew and Greek first. A claim that two words are the same word gets checked in the original language before it's written.
• Every AI edit to the book is logged.

How You Use It

• Desktop app: a search screen with study cards, a reader for browsing the library, dictionary and Strong's lookups, saved searches, and a study to-do list.
• Command line: the same search engine, for scripting and for AI sessions.
• Local web service: the same search over a web connection, for other programs.

Built with Python and SQLite, all running locally.

What I've written here reflects my own reading and conclusions, and I may well be wrong in places. That's why every source and calculation is laid out for you: so you can test the evidence and come to your own conclusions.