AI Amidst the Pews

"These were more noble than those in Thessalonica, in that they received the word with all readiness of mind, and searched the scriptures daily, whether those things were so." — Acts 17:11

This week a colleague mentioned, in passing, that her pastor had preached a mini-sermon on the danger of AI. The thesis: AI does not always get it right.

Pastor, you are correct. It does not. It often answers with unearned confidence, it fills gaps with fluent guesses, it flatters the person asking, and it will occasionally invent a citation with a straight face. Every word of that warning is true, and I would not retract one syllable of it.

Now consider what the warning actually establishes. 'Does not always get it right' is not a property of the machine. It is a property of everything that has ever stood between a reader and the text — every commentary, every study note, every translation committee, every sermon, including the one just preached. The church has never once possessed an infallible intermediary, and it has known this about itself since the 1st century, because the 1st century is when the remedy was recorded. When Paul himself — an apostle, speaking with more authority than any tool or teacher since — preached at Berea, the hearers were not commended for swallowing it. They were commended for checking it: searched the scriptures daily, whether those things were so. Fallibility is not the argument against the tool. Fallibility is the argument for the protocol. The question in front of every shepherd right now is not whether the new intermediary errs. It errs. The question is whether your house has a process for error — because your people are already asking it questions, and the only thing your sermon decided is whether they bring the answers to you.

The Fear, Named Honestly

Let me name what is actually feared, because the sermon about accuracy is rarely about accuracy.

The fear is displacement. For as long as anyone alive can remember, the pastor held a de facto monopoly in his congregation: he was the one who had read the languages, weighed the commentaries, and could answer the hard question after the service. Now a machine answers the hard question in the car on the way home, in depth, with cross-references, at midnight, without appointment. The fear is that the office of teacher is being made redundant — and underneath that, a truer and more honorable fear: that the sheep will be fed error at scale, by a voice with no ordination, no accountability, and no love for them.

Both fears are legitimate. Neither is answered by the bunker. Here is the arithmetic of quenching: the curious do not stop being curious when you shush them. They stop being yours. Every question ruled out of order in the fellowship hall gets asked an hour later in the kitchen, alone, with no elder in the room — which is the exact condition under which error propagates best. The insular church does not protect its people from the tool. It guarantees they meet the tool unaccompanied. Amos saw the end state of a people seeking the word without finding a table set for the search: 'they shall run to and fro to seek the word of the LORD, and shall not find it.' The running to and fro has begun. The only open question is whether it runs through your doors or around them.

3 Eras of Access

This has happened before, and the church's own history is the best teacher in the room.

For roughly a thousand years, the scriptures lived in the memorization era. Access meant a life given to it: men and women in the monastic houses who copied the text by hand and carried whole psalters by heart — the word written, as Paul put it, 'not in tables of stone, but in fleshy tables of the heart.' The scholarship was deep and the doorway was narrow: to search the scriptures, you more or less had to become one of the few whose full-time vocation was to hold them.

Then, around 1525, the doorway blew open. Print, and the vernacular translations, and soon the concordances — and suddenly a plowboy could own the text and search it. Mark carefully what the established pulpits of that day did: many fought it. They burned the translations and sometimes the translators, on an argument that should sound familiar — the untrained reader will get it wrong. And the untrained reader did get it wrong, regularly. The churches that answered the errors by fighting access lost the argument, and very nearly lost the congregation. The churches that answered by catechizing the reading — building the plowboy a bench, a method, and a teacher who welcomed the question — became the engine of the most scripturally literate generations in history. Access was never the danger. Unaccompanied access was.

Now, since roughly 2025, a third doorway: the compute era. One sincere student of the text — one, sitting at a kitchen table — can now overlay the entire manuscript record in every language, the lexicons, the concordances, and alongside them the whole preserved corpus of history, science, and cultural memory, and ask questions across the composite. In candor: the ability to run inference over substantially everything knowable is novel in the whole of recorded history. What print did to reading, this does to searching. And the pattern of the last doorway is repeating in front of us, pulpit by pulpit, with the same 2 responses on offer and the same 2 outcomes waiting.

What the Tool Actually Is

Precision matters here, because both the panic and the hype get the object wrong.

The tool is not an oracle. It is not a spirit, and it is owed none of the tests or courtesies of one — though John's command covers it neatly anyway: 'believe not every spirit, but try the spirits whether they are of God.' It has no ordination, no anointing, and no standing to settle anything. What it is, honestly described, is the most powerful concordance ever built, with a conversation for an interface: a research assistant of unlimited patience, considerable range, and imperfect honesty. It can hold every occurrence of a Hebrew root in view at once. It can also invent a source. Both facts are load-bearing.

And here is what the fearful sermon leaves out: the same class of tool now cracking mathematics problems that stood unsolved for decades can be pointed at the oldest interpretive discipline the church possesses — scripture interpreting scripture — and run it at a scale no paper concordance ever allowed. Every usage, every echo, every translation choice, laid side by side in seconds, for a layman. The tool does not make the finding true; it makes the checking cheap. That is the honest claim, and it is enough: verification, which used to cost a seminary education, now costs a question. A generation of Bereans just had their daily searching subsidized. The only question is whether anyone teaches them to search like Bereans — because the same subsidy, unaccompanied, funds the other thing, and Paul named the other thing in advance: 'ever learning, and never able to come to the knowledge of the truth.'

The Berean Protocol — 5 Marks

So here is the framework, and notice that not one plank of it is new. It is the 1st-century process, restated for the tool. A church that adopts it does not need to fear a single question its people bring home.

1: Everything enters as hypothesis Output from the tool arrives at the bottom of the ladder of authority — below the sermon, below the commentary, below the creed, and infinitely below the text. It is a lead, not a finding. The Bereans extended exactly this discipline to Paul; your people can extend it to a machine. The moment a chat transcript is quoted in the fellowship hall with the cadence of a proof-text, the protocol has failed.

2: Reconcile to scripture Let scripture interpret scripture. Every lead lives or dies by the text — and for the first time in history the layman can actually run the check: the lexicons, the interlinears, the usage tables are one tap away, and the tool itself will walk him through them if asked. The rule of the house: never accept the tool's summary of a verse when the verse itself is free. Open the passage. Read the context. Chase the root. The tool proposes; the canon disposes.

3: Mark your tiers out loud Teach your people to say plainly, every time: this is shown, this is witnessed, this is believed. A finding reconciled to the text and checkable by anyone is one thing; a pattern noticed but not yet proven is another; a conviction held by faith is a third — and all three are honorable exactly as long as they are labeled. A teacher who blurs what he can show with what he only believes has started charging admission. This single habit — tier-marking — converts the tool's flood of suggestions from a menace into a sorted inbox.

4: Verify in the body Findings come to the table, not to a private feed. The insight that cannot survive the room — the elders, the languages, the awkward question from the back row — was not ready, and the one that can survive the room just made the whole body richer. Private revelation, privately confirmed by a machine that flatters its user, is the precise failure mode of this era; the congregation is the corrective. This is what the church uniquely offers that no tool can: a room full of people who love you enough to tell you no.

5: The shepherd referees the method, not the access The gatekeeping of information is over; it ended the way it ended in 1525, permanently. What remains — larger than what was lost — is the gatekeeping of process: the pastor as the one who trains the congregation to check, who models the tier-marking, who hosts the table where findings are tried. 'Every scribe which is instructed unto the kingdom of heaven... bringeth forth out of his treasure things new and old.' That is the job description now. Not the last man with the answer — the first man with the method.

The 2 Ditches

The road has a ditch on each side, and it is worth staring into both.

The first ditch is the one the fearful sermon rightly dreads: inquiry unmoored. The seeker who treats the tool as oracle, collects epiphanies the way some men collect debts, reconciles nothing, submits nothing to the body, and mistakes the pleasure of novelty for the arrival of truth — 'ever learning, and never able to come to the knowledge of the truth.' This ditch is real, it is already populated, and nothing in this letter minimizes it. But mark its cause: it is not caused by access. It is caused by access without protocol — the plowboy with a Bible and no catechism, updated 500 years.

The second ditch is the bunker, and it is deeper than it looks. The church that answers the era by prohibition does not produce a congregation without questions; it produces a congregation that asks elsewhere. The famine Amos described is not a shortage of appetite — the running to and fro is the appetite — it is a shortage of tables. Rule the search out of order, and the search relocates: to the feed, to the forum, to the midnight session with a machine trained to agree, with no elder present and no room to survive. The bunker does not protect the flock from the first ditch. It delivers them to it, one private conversation at a time, and then wonders where the young families went.

Between the ditches runs the old road: 'prove all things; hold fast that which is good.' Not gullibility, not prohibition — proving. The command assumes the encountering. It only asks that the encountering come home to be tried.

What This Looks Like on a Wednesday Night

None of this requires a committee, a budget, or a seminary. It requires one evening and a shepherd willing to host it. Concretely:

Run the inoculation drill first. Before anything else, spend 20 minutes catching the tool in an error, together, on purpose — ask it something obscure enough that it invents, then trace the invention against the text and watch it dissolve. Nothing teaches 'hypothesis, not oracle' faster than a room full of people who have personally watched the confident answer fail the check. Do this once and the naive awe is gone forever; the tool shrinks to its true size, which is still large.

Then teach the check itself. One passage, one evening: take a claim the tool offers about a verse, open the interlinear, chase the root word through its other occurrences, read the surrounding chapter, and render the verdict as a room. Your people will discover the lexicons are no longer locked behind Greek 101 — and that the discipline is a joy, which is the best-kept secret of the whole affair: 'it is the glory of God to conceal a thing: but the honour of kings is to search out a matter.' The concealing was never stinginess. It was an invitation, and the search was always meant to be the fellowship.

Then set the house rules in a paragraph, not a policy binder: findings come to the table; transcripts are leads, not proofs; tiers get marked out loud; the text outranks everything; and no question is out of order — only unfinished. Post it, and mean it.

A fuller kit — the drill scripted, the check-workflows written out step by step, prompt patterns that force sources and steelman the opposing reading, a template for tier-marking, and answers to the objections your elder board will raise — is being assembled at thetimeofthereturn.com/toolkit, free, for exactly this Wednesday night.

So — the Open Door

Step back and look at what is actually on offer.

Every town in the country now contains people running inference across the whole of the knowable, some of them on the scriptures, most of them alone. They are going to keep doing it; that ship has sailed and is not returning to port. The only unsettled question is where the findings get tried — and the conventional church, the one with walls and elders and a Wednesday night, is sitting on the exact infrastructure the moment requires: a body, a method 20 centuries old, and a Book that has absorbed every honest question ever put to it and stands unbothered. The text is not fragile. It never was. The institution that treats it as fragile teaches the flock more about the institution's confidence than about the text.

The church that builds the Berean bench does not merely survive the AI era. It becomes what it was in Berea: the place in town where truth-claims come to be tested, where the curious are welcomed instead of managed, where the searching that Amos saw running to and fro finds a table with its name on it. Faith-enhancing scholarship as the house specialty — not despite the acceleration, but riding it. The alternative was tried in 1525, at scale, with every advantage, and it failed against a printing press. It will not do better against this.

Pastor, your sermon was half right. Finish it. The danger is real, the protocol is older than the danger, and the same congregation you warned on Sunday could be searching the scriptures daily by Wednesday — whether those things are so. That is not the threat. That is the commendation, waiting to be earned again.

---

House rules, as always. The historical claims above are checkable — the monastic transmission of the text, the vernacular-Bible conflicts of the early print era and their outcome, the food-shortage-of-tables reading of the present moment — and the toolkit at thetimeofthereturn.com/toolkit will carry the sources and the scripted exercises. 2 framings are marked as this house's convictions, not as proof, and are argued elsewhere on the site rather than assumed here: the reading of the 525–1525 monastic millennium within the larger prophetic calendar, and the description of the present compute era as a thinning of the veil. A letter to shepherds should not smuggle its eschatology in the mail; the argument above stands on plain history and Acts 17:11 alone, and asks nothing of the reader but the Berean's readiness. Consistent-with is never proven-by. Prove all things; hold fast that which is good.