On Monday Twitter broke for the sixth time this year. Clicking any link on the social network resulted in an error message, while attempting to post a new image resulted in nothing but a big blank box where the picture should have been.
Unlike the last four outages – three in February, and another already in March – the site wasn’t completely unavailable, giving Twitter users the opportunity to engage in their favourite activity: discussing the continued destruction of Twitter live on the site. (The sixth outage this year, in January, only affected Android users.) The trending topics on the site were promptly filled with various phrases relating to the outage, as users speculated that Elon Musk’s own demands had ultimately caused the failure.
The error message referred to the Twitter API, a service that allows other programmes to interact with the site. In January, the company had, with no notice, banned third-party apps from using the API, rendering them non-functional overnight. The error message suggested that Twitter had continued tinkering with the feature and accidentally banned its own apps in the process.
“A small API change had massive ramifications,” Musk confirmed later that day, calling the code of the site “extremely brittle for no good reason”. Fixing the problems, he added, would “ultimately need a complete rewrite”.
Just a single reliability engineer had been allocated to a project to create a paid version of the API, according to a report in the Platformer newsletter, and that engineer pushed a mistaken change to the live version of the site without realising the ramifications.
Mistakes happen, and Twitter isn’t alone in being knocked offline by a single engineer: in 2021, Facebook’s services were offline for almost six hours after the company accidentally deleted itself from the “phone book” of the internet. But the regularity of the outages at Twitter, which seems to be getting worse over time, has left some wondering if the occasional errors are adding up to a more systemic problem.
Steven Murdoch, a professor of security engineering at University College London, said substantial reductions in Twitter’s engineering team meant fewer people to monitor systems and catch minor problems before they become bigger.
“The result is that eventually problems occur which are sufficiently serious to be noticed by large portions of Twitter’s user base,” he said. “It has been suggested that rewriting the Twitter code could help avoid such problems but this is a high-risk strategy. It would result in splitting the already-reduced engineering team between maintaining the old code and making the new version.”
Pushing for such a rewrite has already been the cause of Musk’s downfall once before. In the early 00s, after his company, X.com, had merged with the creators of PayPal, Musk as chief executive led a project to merge the back-ends of the two companies, which had been built on incompatible systems. The rewrite spiralled into “a holy war”, according to a 2007 profile of his time at PayPal, and just six months after the merger, Musk was fired as chief executive while on honeymoon.
At Twitter, a rewrite will also have to compete with other demands on the company’s time. On Tuesday, the Financial Times reported that the EU had demanded Musk hire more moderation experts in order to comply with the bloc’s new regime for regulating social media platforms, while the US Federal Trade Commission is investigating the company amid concerns that Twitter’s ability to protect users might have been affected by redundancies. Since Musk completed the $44bn (£37bn) takeover of Twitter in October last year, the workforce has been reduced from 7,500 employees to about 2,000.
On Tuesday, Musk apologised to a Twitter worker after becoming engaged in a spat with him over whether he still worked at the company or not. The exchange included Musk referring to Haraldur Thorleifsson’s disability as an “excuse” for his performance at work. Musk, who later apologised to Thorleifsson over a “misunderstanding” that led to his tweets, said Thorleifsson was now considering staying at the company.
Cost-cutting measures under which Twitter has stopped paying bills for office space, janitorial supplies and even web hosting are starting to meet pushback. The company stepped back from a bruising confrontation with Amazon Web Services as it tried to negotiate a discount on its cloud computing, according to the news site The Information, while a similar standoff with Salesforce saw its Slack communications platform disabled entirely over one weekend in February, leaving staffers unable to collaborate.
Twitter’s engineering problems are taking place against a backdrop of financial pressure against the company, which took on $13bn of debt as part of Musk’s funding package for the takeover. That debt requires quarterly interest payments of about $300m.
Musk made the first payment in January but analysts have warned that the business, which was loss-making for the majority of its existence before the takeover, must stage a turnaround in order to make those payments consistently. The Tesla boss said this week that Twitter had a “shot” at being cashflow positive in the next financial quarter, referring to one of the financial metrics that can be used to indicate whether a company can meet its debt payments. But at what cost? Twitter has a “shot” at being financially sustainable due to cuts that appear to be affecting the reliability of the site as well as alarming regulators.
According to Twitter’s last set of quarterly results as a listed company, published in 2022, Twitter generated negative free cashflow (spending more cash to run the business than it takes in) of nearly $124m.
Musk also reiterated his plans to introduce payments on Twitter, and said he envisions users could eventually send money to each other with one click.
“I think it’s possible to become the biggest financial institution in the world,” he said.
However, advertising is Twitter’s largest source of revenue, accounting for the majority of its $5.1bn turnover in 2021, and Musk admitted soon after the $44bn takeover in October last year that there had been a “massive” drop in advertising revenue. In January it was reported that Twitter’s daily revenue was down 40% compared with the previous year.
Farhad Divecha, the managing director of UK digital marketing agency Accuracast, said: “Trust is a big factor right now, and these outages do not instil a lot of confidence in Twitter’s ability to service our clients to the level where we or the brands we represent feel comfortable putting money in the service.”
Site breakages add another complication to Musk’s drive to put Twitter on a sustainable footing.
