Read on Twitter
Every day at 00:00 pacific time, a microservice running in a container on a cloud platform rolls over a counter. That counter triggers a notification to another microservice, which creates a new row in a database.
At the same time, another service queries a database. It is returned a result set of around 145 million rows.
Depending on local conditions and hundreds of environmental variables, between sixty and five thousand further services compute scores based on the hundreds of millions of records keyed to each of the 145 million rows.
It can take up to 10 hours for this process to iterate through to a final row with a stable computed score.
An identical version of this process is simultaneously executed in Iceland, France, Australia and Taiwan.
All five results must be identical.
All five results must be identical.
The result is seeded into the timelines of select individuals, whose selection is also computed daily in a similar process.
Prior to seeding, 4096 watchdog daemons are spun up and provided access to shard history.
Prior to seeding, 4096 watchdog daemons are spun up and provided access to shard history.
There is one, singular daemon whose only purpose is to watch for the seeding process and ensure the result is injected into the timeline. We understand the name for this daemon is SHOWRUNNER.
If the result hasn’t been injected by 12:00 pacific, SHOWRUNNER begins shutdown.
If the result hasn’t been injected by 12:00 pacific, SHOWRUNNER begins shutdown.
We are led to believe this entire system — the selection algorithms, services, seeding process and the SHOWRUNNER daemon — is known as MAINCHAR.
MAINCHAR might sound innocuous - surely just surfacing engaging content? - but is nothing of the sort.
MAINCHAR is an algorithm for blood sacrifice, identifying an account to be thrown into the timeline for judgment, in order to keep the machinery of twitter running.
MAINCHAR is an algorithm for blood sacrifice, identifying an account to be thrown into the timeline for judgment, in order to keep the machinery of twitter running.
There is no stopping MAINCHAR.
It is unclear why MAINCHAR is required of Twitter and not other social networks.
Extensive research conducted on Facebook — before researchers were unceremoniously barred — found scant evidence for MAINCHAR there.
It is unclear why MAINCHAR is required of Twitter and not other social networks.
Extensive research conducted on Facebook — before researchers were unceremoniously barred — found scant evidence for MAINCHAR there.
During 2020, there were faint signals indicating potential presence of *multiple* instances of MAINCHAR operating at the hyper-local level on Nextdoor.
Historians of science, technology and media argue that MAINCHAR is not a unique, internet development and that there is non need (in this case) to fear the algorithm.
They point to an earlier, analogue version of MAINCHAR developed in the UK media and found in ancient CMSes known as BLEED_LEAD.
BLEED_LEAD likely worked on a much slower cycle. BLEED_LEAD might have iterated over two, three or even five days or more.
BLEED_LEAD likely worked on a much slower cycle. BLEED_LEAD might have iterated over two, three or even five days or more.
It is 9:46pm on the west coast.
In just over 7980000 milliseconds, MAINCHAR will iterate, again. Just as it has done every single day for over the last ten years.
MAINCHAR will return a result, again.
MAINCHAR will provide a sacrifice, again.
In just over 7980000 milliseconds, MAINCHAR will iterate, again. Just as it has done every single day for over the last ten years.
MAINCHAR will return a result, again.
MAINCHAR will provide a sacrifice, again.
Because the most important thing to understand is this:
MAINCHAR is an automated system.
No humans are involved in its operation or maintenance. No humans are involved in its selection process.
MAINCHAR is an automated system.
No humans are involved in its operation or maintenance. No humans are involved in its selection process.
In 2010, despite extensive effort, including the contracting of deniable librarian and information science assets, Twitter’s leadership were unable to find any documentation for MAINCHAR.
So in 2011, a team from Netflix were deployed to Twitter with a single objective: document MAINCHAR, with a goal to decommissioning.
They failed.
They mitigated MAINCHAR by delaying its selection routine. That method, still used today offensively at Twitter, and defensively elsewhere, is known as CHAOSMONKEY.
CHAOSMONKEY is the *only* reason why it now takes MAINCHAR up to 10 hours to return a result.
They mitigated MAINCHAR by delaying its selection routine. That method, still used today offensively at Twitter, and defensively elsewhere, is known as CHAOSMONKEY.
CHAOSMONKEY is the *only* reason why it now takes MAINCHAR up to 10 hours to return a result.
By 2015, the European Council had become aware of MAINCHAR by way of backchannel.
Twitter and its close allies remained focused on a software solution to MAINCHAR. But Eurocrats tied MAINCHAR to disturbing developments in the intentional formation of an attention economy.
Twitter and its close allies remained focused on a software solution to MAINCHAR. But Eurocrats tied MAINCHAR to disturbing developments in the intentional formation of an attention economy.
Europe’s technocratic leaders quickly developed a policy-based solution to the problem of MAINCHAR.
It became known as the Right to be Forgotten, the extreme measure of complete data erasure.
MAINCHAR — as yet — has not returned a result that is not an active account.
It became known as the Right to be Forgotten, the extreme measure of complete data erasure.
MAINCHAR — as yet — has not returned a result that is not an active account.
Today, MAINCHAR iterates over the ~145 daily active users on Twitter.
But platforms grow and change. Environments are dynamic and respond.
The accounts from which MAINCHAR produces a selection of course react to each day’s MAINCHAR result.
But platforms grow and change. Environments are dynamic and respond.
The accounts from which MAINCHAR produces a selection of course react to each day’s MAINCHAR result.
Forensic examinations of MAINCHAR’s systems are incomplete. Much of MAINCHAR’s precise workings remain unknown.
But there are two key points to consider:
But there are two key points to consider:
1) Processing power continues to increase as processing cost continues to decrease, all while methods for analyzing large datasets — including AI/ML, however accurate or dubious — are refined and improved;
2) The audience to which each day’s MAINCHAR result is provided is insatiable, savage and quick.
This leads us to a hypothesis: the audience may be able to sustain *two* MAINCHAR results in a single 24 hour MAINCHAR period.
As far as we are aware, there are no public systems or organizations monitoring for movement in such a direction.
As far as we are aware, there are no public systems or organizations monitoring for movement in such a direction.
No two-instance MAINCHAR warning system exists or is planned.
It is 10:28pm on the west coast.
In just over 4860000 milliseconds, MAINCHAR will run again.
In just over 4860000 milliseconds, MAINCHAR will run again.
MAINCHAR is coming.
It cannot be stopped.
It cannot be reasoned with.
Just hope that MAINCHAR does not return you.
It cannot be stopped.
It cannot be reasoned with.
Just hope that MAINCHAR does not return you.
~2100000 milliseconds to MAINCHAR selection initialization.
