The Fork Nobody Is Framing Right

People keep asking if Bitcoin will get a quantum fork. That is the wrong question.

The real fight is simpler and much harder. Will Bitcoin upgrade itself through a soft fork, or will the fight over already exposed coins force a hard split? One path keeps one chain. The other path makes two.

I think it will be the fork. If I am right, you want your Bitcoin in place before that split, not after the headlines catch up.

What just happened

On August 26, 2026, StarkWare researcher Avihu Levy got the first quantum-safe Bitcoin transaction mined on mainnet. Block 964199. Ten thousand satoshis. No soft fork. No consensus change.

The script is nonstandard, so it went through MARA Slipstream instead of the public mempool. The cost is a few hundred dollars in GPU time per transaction.

That is a lifeboat, not a fleet. Even StarkWare's CEO said the real answer is still a protocol upgrade.

What the demo proved is narrower and more important. You can move coins into storage a quantum computer cannot open, today, if you know the trick exists and you are willing to pay for it.

Most holders do not know. Most never will until it is late.

The proposals that actually matter

BIP 360 adds a new output type called Pay to Merkle Root, or P2MR. It works like Taproot with the dangerous part removed. The key path a quantum computer can attack is gone. The proposal has been in the official BIP repo since February 2026. That is the upgrade rail.

BIP 361 is the fight. It is a phased sunset of legacy signatures. After activation, Phase A stops new deposits to vulnerable address types in about 3 years. Phase B, about 5 years after activation, stops spending those old signatures at the consensus layer.

Coins that never moved get frozen unless a later recovery path lands.

That is not a technical footnote. That is property rights meeting physics.

The pile sitting in the open

About 6.5 to 7 million BTC already sit in addresses where the public key is visible on chain. That is roughly a third of supply.

Some of that is address reuse by living holders and exchanges. Some of it is ancient Pay to Public Key output, including the dormant pile widely attributed to Satoshi.

Freeze those coins to keep a future quantum thief from spending them. Or leave them spendable and accept that a machine might take them.

There is no third option that keeps every old rule and every old coin untouched.

Standalone experiments exist. QSBitcoin. Q-BTC. Other post-quantum forks. They are useful labs. They are not the market. The market will decide which chain is Bitcoin after a split, the same way it decided after 2017.

Two paths

Path one. Bitcoin activates P2MR, then a careful migration. Wallets, exchanges, and custodians move. The chain stays one.

Path two. The freeze becomes the hill. One camp will not confiscate lost coins, even to stop a future thief. The other camp will not leave a third of supply sitting as a treasure map. Two chains. One froze the exposed coins. One did not.

Committees prefer path one. History favors path two when the moral claim is that strong.

The clock everyone is using is wrong

The published clock assumes a straight line. BIP 361's own phases run about 5 years after activation. People sketch a first soft fork in 2027-2029 and a hard split in the early 2030s. Google's public quantum roadmap still gets talked about around 2032 for a cryptographically relevant machine.

That clock ignores AI.

Google's quantum work leans on classical simulation, error-correction design, and search through huge design spaces. That is exactly the work AI compresses. AlphaFold did not wait for more wet labs. It collapsed a decade of structure work into months.

The same pattern is already running on quantum hardware. Material candidates. Qubit calibration. Error-correction codes written by machines instead of by hand. Even fabrication is starting to compress through better process search.

The bottleneck is not a mysterious wall called physics. The bottleneck is how fast intelligence can search the design space. That curve is exponential, not linear.

So 2032 is not a floor. It is a ceiling the industry is already pushing on. The fork window that matters is the next 2 to 3 years, not the next decade.

The asymmetry

Bitcoin's defenders move at committee speed. BIPs. Review. Miner signaling. Wallet vendors. Exchange legal teams. That is the design. It is also the risk.

The threat moves at AI speed.

By the time a freeze activates on the published calendar, the machine those rules were written to outrun may already exist. The 6.5 million exposed coins will still be sitting there. The StarkWare lifeboat will still only protect the people who knew to climb in.

This is not only a cryptography problem. It is a migration-before-capability race. Almost nobody is framing it that way.

What to do with this

Do not wait for a press cycle to tell you the split is here.

Get coins you care about off reused addresses and off old exposed formats. Be ready to move when the new rail is real. Write your own thesis down while the room is still quiet.

Take the open door. The open door is preparation before consensus. The closed door is arguing on social media after the first honest quantum scare.

If the upgrade stays one chain, you lose almost nothing by being early. If it splits, being late is the whole loss.