Twitter death watch captivates millions




Comment

SAN FRANCISCO — New Twitter owner Elon Musk summoned all employees who “actually write software” to a Friday afternoon summit at the company’s San Francisco headquarters as he plots next moves amid a chaotic exodus of hundreds of employees.

He also tweeted new content moderation policies, saying that hate speech would see its distribution limited and that a handful of controversial accounts had been restored, but not that of former president Donald Trump.

The series of emails and tweets underscored the ad hoc nature of Musk’s oversight of a social media platform that has 237 million daily users, many of whom tweeted their concerns in the last 24 hours that Twitter is about to collapse — something experts said was a likely eventuality, though not necessarily imminent.

Musk’s emails to engineers, however, suggested concern inside Twitter that a crisis was in the making. The emails even went out to staffers who’d walked out Thursday rather than sign a pledge to work “long hours at high intensity.”

Musk also asked all recipients to send him screenshots of their recent code and explain what it had accomplished. While the initial email included no instructions for remote employees, a subsequent one said Musk would try to speak via video to them — but that employees were only excused if they have a family emergency or “cannot physically get to Twitter HQ.”

In a third email eight minutes later, he asked employees to fly to San Francisco, saying he would be at the office until midnight on Friday and back again Saturday morning. Yet another missive an hour later said flying “would be appreciated, but is not essential.”

The requests to help him “better understand the Twitter tech stack” struck many engineers as absurdly late, given that he had fired about half of what had been more than 7,500 employees two weeks ago and then issued an ultimatum on Wednesday that prompted a subsequent wave of departures.

Several critical teams essential to keeping the site functioning were cut to a single engineer or none by the departures Thursday, leaving the company partially on autopilot and likely to crash sooner or later, engineers said.

“Every mistake in code and operations is now deadly” said a former engineer who departed the company this week. Those left “are going to be overwhelmed, overworked and, because of that, more likely to make mistakes.”

The team that runs the service Gizmoduck, which powers and stores all information in user profiles across the site, was entirely gone, according to a recent department head who spoke on the condition of anonymity because he was not authorized to detail the departures.

Even without a mistake, the system can only run for so long with so little upkeep, tweeted Ramin Khatibi, a site reliability engineer who left Twitter in 2019: “The fact that Twitter continues to work is a testament to the 1000s of engineer years spent building that reliability. But as engineers, we know that failure is coming without continued investment to protect against the next thing.”

It wasn’t clear who was still employed and for how long. Access to internal systems had not been cut for those who walked out, and it wasn’t certain to all that Musk would fulfill his promise to give three months’ severance pay to those who declined to sign a pledge that called for working longer hours. Employees estimated that roughly 1,000 refused to sign the pledge Thursday.

Half the trust and safety policy team resigned, including a majority of those who work on spotting misinformation, spam, fake accounts and impersonation, according to two employees familiar with the team.

Yet Musk plowed ahead with moderation decisions, announcing on Twitter that the company had reinstated accounts including those of comedian Kathy Griffin, who had mocked him, conservative humor site Babylon Bee, and right-wing self help guru Jordan Peterson. Musk had previously said banning and unbanning would be decided with input from a broad council of outside advisers.

He also said that hateful or negative tweets would be barred from algorithmic boosting, even though the team working on ethical artificial intelligence had been laid off.

In an opinion piece Friday in the New York Times, former head of trust and safety Yoel Roth said he had resigned recently because it was clear Musk’s capriciousness would continue, making already complex decisions about content unworkable.

“A Twitter whose policies are defined by unilateral edict has little need for a trust and safety function dedicated to its principled development,” Roth wrote.

Twitter’s systems are complex, and it wasn’t clear if there would be immediate technical failures, said an employee who did not sign Musk’s pledge and spoke on the condition of anonymity to candidly discuss the state of the platform. The systems could suffer during the World Cup, typically a time when Twitter use surges.

The systems could also slowly degrade over time, the person said. Or it could take longer to make fixes to routine problems that arise following the departures.

Most risky of all, this person and another former engineer said, will be when staff try to implement one of the many new things Musk wants.

“Complex systems such as Twitter are most in danger of breaking when engineers attempt to make live changes to add features,” the second person said. “Any change is inherently risky. Twitter is a global communications utility. Change at our scale means even the smallest issues have huge impact.”

Will Oremus and Faiz Siddiqui contributed to this report.