Terrorist played social media firms to a tee, and used them and famous commentators to spread his message of hate
It was just before 2pm local time when Brenton Tarrant opened fire at a mosque in Christchurch, New Zealand.
Strapped to his head was a camera that live-streamed every terrifying moment of his attack to millions of people around the world.
That camera was, arguably, one of his most powerful weapons. This was a man who knew how to exploit the online system and did so to devastating effect. With his live-stream, he used social media to spread his hate-filled views while the world watched on from behind the safety of their screens.
Within seconds, footage of the murders, along with his manifesto, spread rapidly across social media sites.
Tarrant had already primed so-called "alt-right" groups online to support him ahead of the massacre, ensuring his action had maximum impact. Posting on the message board 8chan, he wrote: "Time to stop s***posting and time to make a real life effort".
Users on sites such as Twitter and Reddit reshared quickly while social media companies tried to take it down. Hours later, around 10 videos were being uploaded to YouTube every hour and were easily searchable.
"I'm watching videos of the incident be uploaded and taken down and then uploaded again in real-time. Truly disturbing," posted writer Ryan Mac.
Artificial intelligence tools used by the likes of Facebook and YouTube to block disturbing content were no match for Tarrant's methodical manipulation of the system.
More than five hours after it was flagged, the entire live stream of the massacre could still be seen on YouTube, without any warnings in place.
Not only did the gunman manage to go viral, but he did so in a targeted way, getting the attention of specific communities, "s***posting" internet jokes and mentioning online celebrities.
S***posting is internet slang for trolling people online with the intention of getting an emotional response and attracting the attention of internet users - and Tarrant was a master. For instance, as Tarrant approached the Al Noor mosque on the live video, he urged his followers to "subscribe to PewDiePie" before he began shooting at his victims.
PewDiePie, whose real name is Felix Kjellberg, is a popular but controversial Swedish internet personality now living in the UK who has encouraged his fans to promote his YouTube channel so that he is not overtaken by another YouTube channel.
Mr Kjellberg distanced himself from the shooting in a post on Twitter to his followers yesterday, writing: "I feel absolutely sickened having my name uttered by this person."
In doing so, however, he drew yet more attention to the video - his 89 million YouTube subscribers helping Tarrant spread his message more widely in a phenomenon known as the "Streisand effect". The Streisand effect is a phenomenon whereby an attempt to hide or censor a piece of information has the unintended consequence of publicising it more widely.
Tarrant's 73-page manifesto also showed an awareness of internet culture and media stereotypes.
In it he joked that "Fortnite trained me to be a killer", a reference to the popular battle royale game seemingly designed to prompt a debate over the violence in the game.
He also singled out Candace Owens, an American conservative commentator, who is outspoken about her anti-Democratic views, as the person who had influenced him the most.
In doing so, he got the attention of Owens's 1.13 million followers on Twitter and 855,000 followers on Facebook.
However, a closer look at the tone in his manifesto when he talks about Owens suggests the reference is simply an attempt to spread political division and get attention.
There were warning signs the shootings would happen, and all were missed by social media companies.
Three days before the shooting, Tarrant meticulously photographed the decorated guns and body armour that he had prepared and shared photographs of them on his Twitter account, knowing the images would be dredged up after the attack.
Tarrant also advertised his Facebook live-stream on 8chan, which has 14.1 million visits per month, encouraging them to follow his actions. Some posted encouraging comments as Tarrant broadcast his killings. Facebook said it had moved quickly to take down the original live stream.
"Police alerted us to a video on Facebook shortly after the live-stream commenced and we quickly removed both the shooter's Facebook and Instagram accounts and the video," the social media platform said.
"We're also removing any praise or support for the crime and the shooter or shooters as soon as we're aware."
Mia Garlick, the director of policy for Facebook in Australia and New Zealand, said the company would "continue working directly with New Zealand Police as their response and investigation continues".
Facebook does have teams in place to try to stop the spread of this type of content, although it has been argued the measures do not go far enough.
The company currently has more than 7,500 reviewers whose job it is to police offensive or graphic footage. Many are paid around $15 an hour to scroll through videos of rapes, torture, abuse and murders.
This, in itself, has led to questions for Facebook, specifically over how it protects those moderators. A lawsuit last year accused the company of "ignoring its duty to provide a safe workplace", something which Facebook said it takes "very seriously".
Facebook has added tools to make it easier for moderators to deal with the videos. They have access to replay functions, timestamps of content reported by users and text transcripts to detect content as fast as possible.
"Until you release live video, you don't fully understand the complexity of what it means to create the right policies for live video," Justin Osofsky, the head of global operations at Facebook, told technology website Motherboard last year. "But you learn. And we learned."
But the latest wave of harmful content suggests they have not learnt enough.
As well as human moderators, social media companies can use image recognition to take down violent or graphic videos. But to block footage in near-real time is beyond current technology says Marc Warner, the chief executive of AI company Faculty - something Tarrant would have been all too aware of.
"Once a video is identified as terrorist to allow people to relatively easily take that down in many places," he says "but it has to be identified first."
While machine-learning and artificial intelligence can be used to search for known videos and delete them, it is harder immediately to roll out this spotting technology to new videos, such as the New Zealand shooting. And, with videos on Facebook Live, this becomes impossible.
Facebook, YouTube, Instagram, Snapchat and other live sharing websites companies have to deal with hundreds of millions of users uploading live footage of themselves all the time.
While some videos may not be seen by many people, because social media companies such as Facebook want to increase engagement, live videos are often promoted ahead of other posts.
This can lead to them gaining comments and other interactions, before the company can get a moderator to take them down.
With social media companies hunting more engagement in this way, policing these videos has become increasingly problematic.
The shocking impact of such footage in this case caused some media outlets to repost the video, spreading it still further. (© Daily Telegraph, London)