What Is a Hash?

In the last days of December 2018, people using a popular desktop Bitcoin wallet started seeing an error message when they tried to send a payment. The message urged them to install an important update. The update was fake. Attackers had set up servers that the wallet connected to, and those servers answered with the warning and a link to a download that looked like the real thing. Over the months that followed, the campaign spread, and the security firm Malwarebytes estimated that it netted the thieves "well over 771 Bitcoins."¹

The people who installed that update had no easy way to tell a genuine file from a convincing copy. Yet there is a way, and it sits next to the download button of almost every serious piece of Bitcoin software: a long string of letters and numbers that most people scroll straight past. It is called a hash, and the same idea holds the entire Bitcoin ledger together.

A Fingerprint for Any Data

A hash function takes any data at all and boils it down to a short fingerprint. A single word, a photo, a whole book: whatever goes in, a fingerprint of the same length comes out. Bitcoin uses a hash function called SHA-256, a standard published by the US National Institute of Standards and Technology.² Written out, a SHA-256 fingerprint is always 64 characters long, using only the digits 0 to 9 and the letters a to f.

Here are three real ones. The second input is Bitcoin's own white paper, the nine-page PDF file published on bitcoin.org.³

A seven-letter word and a nine-page document come out exactly the same length. And the process is completely repeatable: run the same data through SHA-256 again and you get the same fingerprint, on any computer, every time. You can check the first one yourself with any SHA-256 calculator, or on a Mac by typing printf Bitcoin | shasum -a 256 into the Terminal app.

Change One Letter

Now look at the last row. The only difference in the input is that the B became lowercase. The fingerprint changed completely. Of its 64 characters, 62 are different, and the two that happen to match are fewer than pure chance would produce on average. Nothing in the new fingerprint hints that the input was almost the same.

That property is the heart of a good hash function. You cannot get closer to a particular fingerprint by getting closer to the data that made it. A near miss looks exactly as random as a wild one, so there is no way to work toward a target step by step. And it is what makes a fingerprint useful for checking a download: change even one tiny piece of a file, a single byte among millions, and its fingerprint looks nothing like the original.

No Way Back

You cannot rebuild a person from their fingerprint, and you cannot rebuild data from its hash. Sixty-four characters simply cannot hold nine pages of text; almost everything about the input is gone, and only the fingerprint is left. The only known way to find data that matches a given fingerprint is to guess, over and over, hashing each guess and comparing. And there are 2²⁵⁶ possible SHA-256 fingerprints, a number 78 digits long.²

In theory two different inputs can share a fingerprint. There are endlessly many possible inputs and only a fixed number of fingerprints, so such pairs, called collisions, must exist. The question is whether anyone can actually find one. For an older hash function, SHA-1, the answer turned out to be yes. In February 2017 researchers at CWI Amsterdam and Google published two different PDF files with the same SHA-1 fingerprint, 38762cf7f55934b34d179ae6a4c80cadccbb7f0a.⁴ Run the same two files through SHA-256 and their fingerprints come out completely different.

As of November 2026, no one has found a collision in SHA-256. The best published research attacks work only on a weakened version with 37 of its 64 internal steps.⁵ That is why the careful phrase is "no known way back," not "impossible." It is a statement about everything anyone has managed so far, and SHA-1 is a reminder that such statements are worth checking now and then.

The Chain Is Made of Fingerprints

Bitcoin runs on these fingerprints. Every block carries the fingerprint of the block before it, which is what turns a pile of blocks into a chain.⁶ Block 974,198, mined on the evening of November 1, 2026, US time, has this fingerprint:

00000000000000000000264584c045b859a3df56321b56202892b67b7971f10f

and the very next block, 974,199, carries that exact string inside it.⁷ Now suppose someone went back and changed one transaction in block 974,198, reducing a real payment in it by a single satoshi, one hundred-millionth of a bitcoin. That transaction's ID, itself a fingerprint, would change. So would the block's summary of all 3,340 transaction IDs it contains. And the block's own fingerprint would become:

b530b762932fe4b0dd18a7a6614dbbd1e2eb7e1e4aeb902a3860efc85b1ea8dd

That is no longer the fingerprint block 974,199 points to, so the link to the next block breaks, and every node would notice at once.⁸ Why that makes the past so hard to rewrite comes down to the other thing worth noticing: the real fingerprint starts with twenty zeros, and the altered one starts with none.

Guessing at the Speed of a Planet

Those zeros are mining. To add a block, a miner hashes the block's header over and over, changing one number in it each time, looking for a fingerprint below a target set by the network's rules. A fingerprint below the target starts with a long row of zeros, and since every attempt is effectively a random draw, the only way to find one is to keep guessing.⁶ Block 974,199's fingerprint starts with 19 zeros. In early November 2026, miners around the world were making about a thousand quintillion of these guesses every second, a 1 followed by 21 zeros.⁹ Someone who wanted to alter an old block would have to redo that guessing for it, and for every block after it, while the rest of the world kept going.

And every Bitcoin transaction's ID is its own fingerprint too, the SHA-256 hash of the transaction's bytes, which is why changing a single satoshi changed the ID above.

Checking the Download

Back to that download button. When Bitcoin Core released version 32.1, its developers published a plain text file called SHA256SUMS listing the fingerprint of each of the 28 files in the release. The fingerprint for the Windows installer, bitcoin-32.1-win64-setup.exe, is:

d4cc64d500fa7713b5e2f3b9d46a215def3c4ed91aa46bf13e4de1a042bf9aa2

After downloading, your computer can work out the fingerprint of the file you actually received. If it matches, character for character, you have exactly the file the developers published. If even one byte had been changed, it would look completely different.¹⁰

There is one more step, and the 2018 attack shows why it matters. A convincing fake site can post a fingerprint of its own fake file. So the list of fingerprints has to be the genuine one too, and Bitcoin Core's developers sign it, so that anyone can check who published it.¹⁰ How a digital signature proves who made something is a story of its own.

So what is a hash? A short fingerprint of any data. The same data always gives the same one. Change anything and it changes completely. And there is no known way back except guessing. That one simple tool lets a stranger check a download, lets every node check every block, and makes Bitcoin's history expensive to rewrite.

Lesson 4 of the course shows how blocks link together and why the record survives. What makes any money trustworthy in the first place is where the free first lesson begins.