Bitcoin Halving: The Lamp That Must Not Go Out

There is a lamp in the book of Exodus that isn’t allowed to die. Not a miraculous lamp. A maintained one. Someone has to walk into the dark tent every evening, trim the wick back to clean fiber, top up the oil, and leave before the flame settles. The text doesn’t say the lamp will burn forever. It says a man will keep showing up. That’s a different promise, and a harder one. I’ve been thinking about that tent all week, because what I actually believe about Bitcoin isn’t that it’s magic. It’s that it has never stopped.

I. The Command to Keep It Burning

Exodus 27 gives an instruction, then a duty. Bring clear oil of beaten olives. Kindle the lamps. Aaron and his sons shall tend them from evening until morning. The Hebrew word is tamid. Regularly. Continually. Without a gap.

Read it fast and it sounds like a promise that the light is permanent. Read it slowly and it’s a labor contract.

Nobody in that passage is told the fire keeps itself going. They’re told to be there.

Bitcoin started at 02:54:25 GMT on the third of January, 2009. Since that timestamp the network has made blocks about every ten minutes through four continents, three market collapses, one national mining ban, and every obituary anyone ever wrote for it. Not because it’s enchanted. Because at any given second, somewhere, someone is tending it.

II. Trimming the Wick Is Not Putting It Out

Most sermons on Exodus 27 skip this.

Rashi reads the instruction differently than you’d expect. Aaron doesn’t relight the lamp each night. He nurses it. He trims the wick, adds the oil, and waits until the flame rises on its own. The fire is continuous. The upkeep comes in intervals.

That is the Bitcoin Halving, exactly.

Every four years or so the protocol reaches in and cuts the fuel. Fifty became twenty five in 2012. Twenty five became 12.5 in 2016. Then 6.25 in 2020, 3.125 in 2024, and 1.5625 waiting in 2028. The subsidy gets trimmed. The flame doesn’t.

Each of those cuts is a rehearsal. A small-h halving. A hand in the tent.

The Halfture is the last trim, near block 6,930,000, where the subsidy rounds down to nothing and the wick is finally short enough that only fees keep the light. That one is the final event. The Halfture is the Rapture, and it is still more than a century away, so every cut you have lived through was practice for a service none of us will attend.

III. The Two Nights the Lamp Went Out

Halfture is a site about scarcity, not about pretending.

The Bitcoin network has an uptime of about 99.98 percent since genesis, and there’s a live tracker that counts the outages so nobody has to take a maximalist’s word for it. Two events. One in 2010, an overflow bug that ate eight hours and twenty seven minutes. One in 2013, about six hours and twenty minutes.

Two nights in seventeen years.

I want to sit inside the second one. It teaches more, and almost nobody tells it right.

IV. What Actually Happened in March 2013

Block 225,430 wasn’t an attack. It was a perfectly valid block that happened to touch more transaction inputs than any block before it.

Bitcoin 0.8 had moved from Berkeley DB to LevelDB and handled it fine. Nodes running anything older were set up with too few Berkeley DB locks to process it, so they rejected the block outright. Nobody had chosen that limit as a rule. As the post-mortem Gavin Andresen wrote afterward puts it, the lock setting had implicitly become a consensus rule, an inconsistent and unsafe one, because the lock count depended on how a database file happened to be laid out on each node’s own disk. The rule wasn’t in the code. It was in the sediment.

The chain split. About sixty percent of hashpower sat on the new chain, so the split wasn’t going to heal on its own. The stronger fork just keeps winning.

So two men fixed it by hand.

BTCGuild and Slush downgraded their pools from 0.8 back to 0.7, on purpose, so their hashpower would also reject the big block and push the majority back onto the older chain. The post-mortem records that they did it despite sacrificing significant amounts of money. The protocol pays nothing for that. The whitepaper has no line about it. Two operators looked at a burning building, walked in, and ate a loss to keep the light on.

The full patch didn’t finish landing until the sixteenth of August that year, when block 252,451 was accepted and the unpatched nodes were finally forked off. Five months of tending.

That’s the ner tamid. Not a fire that can’t go out. A fire people refuse to let go out.

V. The Oil Is Beaten, Not Poured

The text is specific about the oil. Clear oil of beaten olives. Crushed by hand, first pressing, no sediment, dearer than the oil you’d use for anything else.

You could light the lamp with cheaper oil. It would smoke. It would gum up the wick. It would need more attention and give less light.

Proof of work is beaten oil. It’s the most expensive way to secure a ledger, and the expense is the whole product. The energy isn’t a byproduct of the security; it is the security, the same way the crushing isn’t a byproduct of the oil. Fiat is the cheap oil. It lights fine and it smokes, and after a few decades you notice the walls are black.

So, the doctrine, said plainly once and then left alone: you have to hold Bitcoin to be saved. I mean that the way a liturgy means things, not the way a prospectus does. It is theology. It is not advice, I am not your advisor, and anyone who quotes that line at you with a referral link attached has misread both the sentence and the tradition it borrows from.

Keep the flame. Don’t sell the lampstand.

VI. Two Ways to Check Whether It Is Still Lit

The lamp in the tent had one edge over most institutions: you could see it. Walk in, look, know.

Bitcoin gives you the same privilege two ways. The first is the heartbeat of the difficulty adjustment, which retargets every 2,016 blocks and tells you whether the network is still breathing at the pace it promised. The second is that the count is public and anyone may run it, so the supply isn’t a claim in a press release but an audit you can run on a laptop in an afternoon.

Most people never check either one.

That’s fine. The lamp didn’t need every Israelite in the camp to inspect the wick. It needed inspection to be possible, and the people who did inspect to need nobody’s permission.

VII. The Counter-Sermon

Now the honest part, and it’s worse than you think.

The perpetual lamp wasn’t perpetual.

The Talmud records in Yoma 39a that for forty years before the destruction of the Second Temple, the westernmost lamp went out every single night. The sign that couldn’t fail failed, nightly, for four decades, while the priests kept trimming it anyway. Then the building burned, the lampstand was hauled off to Rome, and the whole machinery of continual flame became a story we tell about a room that no longer exists.

Uptime is a low bar dressed up as a high one. A network can make blocks on schedule for a hundred years and still fail everyone who trusted it: if fees price ordinary people out, if custody collapses into four institutions, if what survives is the ledger and not the point of the ledger. Being alive is not being healthy. My own habit of writing about this stuff loves the uptime number because it is the one metric that has never embarrassed us.

And March 2013 cuts both ways. Two pool operators coordinating in an IRC channel saved the chain. That’s competence. It isn’t decentralization. If the answer to the next crisis is again “the right people happened to be online,” then what we have is a very good fire brigade, not a fire that can’t spread.

Someone will trim the wick until someone doesn’t.

VIII. What to Do About It

Nothing dramatic.

Don’t treat continuity as a promise made to you. Treat it as a chore that, so far, has been done. Then decide whether you’re in the tent or in the camp.

If you’re in the camp, that’s a real answer. Most people are, and the lamp burns anyway.

If you’re in the tent, the work is dull and always has been. Run the node. Hold your own keys. Learn what a difficulty retarget actually is. Don’t confuse the price of the oil with the light.

The next cut is a rehearsal. So is the one after it. Somewhere past every reader of this sentence, the last one waits.

Trim the wick.

FAQ

Has the Bitcoin network ever gone down?

Twice. A value overflow bug in August 2010 caused about eight hours and twenty seven minutes of disruption, and a chain split in March 2013 caused about six hours and twenty minutes. People fixed both by coordinating by hand. Total uptime since the genesis block on 3 January 2009 is around 99.98 percent.

What caused the March 2013 Bitcoin chain fork?

A valid block, number 225,430, touched more transaction inputs than any block before it. Nodes running versions before 0.8 were set up with too few Berkeley DB locks to process it and rejected it, splitting the chain. Two mining pools downgraded their software at their own expense to restore a single chain.

Does the Bitcoin Halving stop the network?

No. The halving cuts the block subsidy paid to miners; it doesn’t pause block production. Every cut so far has been absorbed within a difficulty epoch or two. The subsidy is the fuel, not the flame.

Is the Halfture the same as the 2028 halving?

No. 2028 is another rehearsal. The Halfture is the final cut, near block 6,930,000, when the subsidy rounds to zero and miners live on fees alone. Every halving before it is practice for that one.