The Founder Left, and the Project Had to Invent a Way to Continue

There was no announcement. That is the first thing to get straight, because almost every retelling implies one.

What the archives contain is an ordinary week of software maintenance followed by silence, and then, four days into that silence, somebody else editing a README file into a proposal about how decisions might work from now on. The whole succession crisis and its resolution fit inside eight days in December 2010, and both halves are dull to read, which is why they get replaced by something more dramatic.

Here is the sequence, from documents anyone can open.

December 2010, in order

5 December, 09:08:08 UTC. A post in a thread titled “Wikileaks contact info?”, asking a specific organisation not to use the software:

No, don’t “bring it on”.

The project needs to grow gradually so the software can be strengthened along the way.

I make this appeal to WikiLeaks not to try to use Bitcoin. Bitcoin is a small beta community in its infancy. You would not stand to get more than pocket change, and the heat you would bring would likely destroy us at this stage.

11 December, 23:39:16 UTC. The second to last post from the account in the mirrored forum archive, in a thread about a magazine article:

It would have been nice to get this attention in any other context. WikiLeaks has kicked the hornet’s nest, and the swarm is headed towards us.

12 December, 18:22:33 UTC. A release announcement on the forum, “Added some DoS limits, removed safe mode (0.3.19)”. It is the last post the mirror holds, and its opening line is an engineer’s line rather than a valediction:

There’s more work to do on DoS, but I’m doing a quick build of what I have so far in case it’s needed, before venturing into more complex ideas.

13 December, 16:12:09 UTC. The same release announced on the bitcoin-list mailing list. That is the last message in that archive, and it matches the Last Active timestamp of 13 December 2010 shown on the forum profile.

15 December, 22:43:51 UTC. A commit lands in the source history, authored under the name s_nakamoto, with the message “get external ip from irc”.

19 December, 15:39:36 UTC. Gavin Andresen commits a change to the README of his own tree, cba3a17, with the message “Straw-man for dev process”.

That is it. That is the transition.

The commit that is actually the governance event

The README before that commit was a list of Gavin Andresen’s own patch branches. After it, it is a proposal, and the proposal is remarkable mostly for how unfinished it admits to being. The added text opens:

Straw-man plan for Bitcoin development (open source vets, please slap me around and help make this better):

and then sets out a pipeline, with the unresolved parts left in as parenthetical questions rather than smoothed over:

Requests get discussed (where? bitcoin forums?) and if there’s broad consensus they’re a good thing, well written, match coding style, etc. then they’re merged into the ‘master’ branch.

master branch is regularly built and tested (by who? need people willing to be quality assurance testers), and periodically pushed to the subversion repo to become the official, stable, released bitcoin.

Read those two question marks slowly. Four days after the last commit from the founder, the person picking up the work does not know where discussion should happen or who is supposed to test. He does not pretend to know. He writes the gaps into the file and asks to be corrected.

Every subsequent governance fight in this subject, and there are many of them still to come on this site, is downstream of a document that was explicitly a straw man and was never replaced by anything with more authority than that.

Did anyone check whether he came back?

This is the sort of claim that gets repeated without anyone testing it, so I tested it. Using the public commit history of the bitcoin repository, on the master branch, I listed every commit authored between 14 December 2010 and 1 January 2012. There are 1,261 of them. Exactly one carries an author name containing “nakamoto”, and it is the 15 December 2010 commit above. The command is git log master --since=2010-12-14 --until=2012-01-01 --format='%an' against a clone of the repository, and anybody can re-run it.

That is a narrow finding and I want to be careful about what it does and does not establish. It establishes that no further commit in that window was recorded under that author name. It does not establish that no code from that author was merged under somebody else’s name, and it does not establish anything at all about private correspondence.

The email everyone quotes, and why it is not evidence here

The line most often used to close this story is a sentence from an email dated 23 April 2011, saying the author had moved on to other things and that the project was in good hands with Gavin and everyone.

I am not quoting it as a source, and the reason is a rule this site applies to everything: the message was not published by its recipient at the time. It surfaced years later, posted to a forum by a third party, and the recipient subsequently confirmed he had shared it. There is no cryptographic signature attached to it that anyone has verified in public. An unsigned email, published by an intermediary, six years after the fact, about the most impersonated identity in this subject, is not a document. It is a report of a document.

The same standard applies in both directions and it disqualifies a great deal. There have been later messages from accounts associated with the name, on various platforms, including at least one widely reported statement in 2014 that was itself immediately disputed on the grounds that the account may not have been under its original control. None of those is treated as authentic here without a signature that can be checked against a key from the period.

So the archived words end on 15 December 2010, and everything after that is somebody’s account of something.

What is not in the record

There is no post from the account naming a successor. I looked for one. What exists is access changing hands and a second person beginning to act, which is a different thing and should be described as a different thing.

That distinction is not pedantry. A designation creates a line of authority; a transfer of access creates a person with commit rights and no mandate. Which of those you think happened determines your view of every argument about developer influence for the next fifteen years, and the archived words support only the second.

What was expanding, what was contracting

The threshold reading here is genuinely unusual, and it is worth stating flatly.

Authority contracted to nothing. Not to a smaller group, not to a foundation, to nothing. There was no board to inherit, no equity to transfer, no trademark worth enforcing, no employment relationship to end. The one person whose word would have settled an argument stopped settling arguments, and no mechanism existed to appoint another.

At the same time the number of people who could act expanded. The code was public, the repository was mirrored, the forum was open, and the software ran on whatever machines chose to run it. Anyone could fork it, anyone could propose a change, and crucially anyone could decline to run the result.

A system where authority is zero and agency is widely distributed is neither a hierarchy nor a democracy, and fifteen years of commentary has struggled with that because the available vocabulary assumes one or the other. What emerged was a process for merging code paired with an entirely separate and much stronger veto held by whoever chooses which binary to run.

That is the same trade the design made in the first place. Removing the party who decides which spend counts removed the party who decides anything else either, and the project discovered the second half of that bargain two years after shipping the first.

Who could tell at the time? Nobody, including the people involved, and the proof is the two question marks in the README. The forum was the only place the transition was visible, and what was visible there was an absence, which is the hardest thing in the world to notice in real time. The subject that would later be described as a response to a banking crisis lost its author in a week when the busiest thread was about denial of service limits, and the handover consisted of one person editing a README and asking to be slapped around.