He could be cryptic, demeaning and scary, sending angry messages and photos of guns. If they didn't respond how he wanted, he sometimes threatened to rape or kidnap them - then laughed it off as some big joke.
But the girls and young women who talked with Salvador Ramos online in the months before he allegedly killed 19 children in an elementary school in Uvalde, Texas, rarely reported him. His threats seemed too vague, several said in interviews with The Washington Post. One teen who reported Ramos on the social app Yubo said nothing happened as a result.
Some also suspected this was just how teen boys talked on the Internet these days - a blend of rage and misogyny so predictable they could barely tell each one apart. One girl, discussing moments when he had been creepy and threatening, said that was just "how online is."
In the aftermath of the deadliest school shooting in a decade, many have asked what more could have been done - how an 18-year-old who'd spewed so much hate to so many on the Web could do so without provoking punishment or raising alarm.
But these threats hadn't been discovered by parents, friends or teachers. They'd been seen by strangers, many of whom had never met him and had found him only through the social messaging and video apps that form the bedrock of modern teen life.
The Washington Post reviewed videos, posts and text messages sent by Ramos and spoke with four young people who'd talked with him online, who spoke on the condition of anonymity for fear of further harassment.
The girls who spoke with The Post lived around the world but met Ramos on Yubo, an app that mixed live-streaming and social networking and had become known as a "Tinder for teens." The Yubo app has been downloaded more than 18 million times in the U.S., including more than 200,000 times last month, according to estimates from the analytics firm Sensor Tower.
On Yubo, people can gather in big real-time chatrooms, known as panels, to talk, type messages and share videos - the digital equivalent of a real-world hangout. Ramos, they said, struck up side conversations with them and followed them onto other platforms, including Instagram, where he could send direct messages whenever he wanted.
But over time they saw a darker side, as he posted images of dead cats, texted them strange messages and joked about sexual assault, they said. In a video from a live Yubo chatroom that listeners had recorded and was reviewed by The Post, Ramos could be heard saying, "Everyone in this world deserves to get raped."
A 16-year-old boy in Austin who said he saw Ramos frequently in Yubo panels, told The Post Ramos frequently made aggressive, sexual comments to young women on the app and sent him a death threat during one panel in January.
"I witnessed him harass girls and threaten them with sexual assault, like rape and kidnapping," said the teen. "It was not like a single occurrence. It was frequent."
He and his friends reported Ramos' account to Yubo for bullying and other infractions dozens of times. He never heard back, he said, and Ramos' account remained active.
Yubo spokeswoman Amy Williams would not say whether the company had previously received reports of abuse related to Ramos' account. "As there is an ongoing and active investigation and because this information concerns a specific individual's data, we are not legally able to share these details publicly at this time," she said in an email. Williams would not say what law prevents the company from commenting.
Texas Republican Gov. Greg Abbott said Wednesday that Ramos had also written, "I'm going to shoot my grandmother" and "I'm going to shoot an elementary school" shortly before the attack in messages on Facebook. And Texas Department of Public Safety officials said Friday that Ramos had discussed buying a gun several times in private chats on Instagram.
Ten days before the shooting, he wrote in one of the messages, "10 more days," according to the official. Another person wrote to him, "Are you going to shoot up a school or something?" to which Ramos responded, "No, stop asking dumb questions. You'll see," the official said.
Andy Stone, a spokesman for Meta, which owns Facebook, Instagram and the chat service WhatsApp, referred The Post to an earlier statement from the company that said the messages were sent privately.
The rise of services that connect strangers through private messaging has strained the conventional "see something, say something" mantra repeated in the decades since the Columbine High School massacre and other attacks, according to social media researchers. And when strangers do suspect something is wrong, they may feel they have limited ways to respond beyond filing a user report into a corporate abyss.
Many of Ramos' threats to assault women, the young women added, barely stood out from the undercurrent of sexism that pervades the Internet - something they said they have fought back against but also come to accept.
A 2021 Pew Research Center study found these experiences are common for young people, with about two-thirds of adults under 30 reporting that they've experienced online harassment. Thirty-three percent of women under 35 say they have been sexually harassed online.
Danielle K. Citron, a law professor at University of Virginia, said women and girls often don't report threats of rape to law enforcement or trusted adults because they have been socialized to feel they do not deserve safety and privacy online. Sometimes, they don't think anyone would help them.
Women and girls have "internalized the view, 'What else do we expect?'" said Citron, the author of the upcoming book "The Fight for Privacy: Protecting Dignity, Identity, and Love in the Digital Age." "Our safety and intimate privacy is something that society doesn't value."
Ramos' hatred toward women and obsession with violence were clear in the messages viewed and interviews conducted by The Post, but his identity was mostly hidden. The teens who spoke with The Post said they saw him on live videos he did on Yubo, then they exchanged Instagram user names to message with him.
And he'd constrained his comments to private messaging services like Yubo and Instagram, leaving only the recipients with the burden to react.
Like many of the people he spoke with, Ramos had shared little about himself online. He used screen names like "salv8dor_" and "TheBiggestOpp" - and shared only his first name and his age. His profile pictures were selfies, him holding up his shirt or looking dour in front of a broken mirror.
He shared animal videos, struck up flirtatious conversations and shared intimate things about his past that left some feeling like distant friends. But in recent months, he'd also started posting darker imagery - moody black-and-white photos and pictures of rifles on his bed.
His threats were often hazy or unspecific, and therefore easily dismissed as just a troll or bad joke. One girl told The Post she first saw Ramos in a Yubo panel telling someone, "Shut up before I shoot you," but figured it was harmless because "kids joke around like that."
In the week before the shooting, Ramos began to hint that something was going to happen on Tuesday to at least three girls, she said. "I'll tell you before 11. It's our little secret," she said he told them multiple times. On the morning of the shooting, he messaged her a photo of two rifles. She responded to ask why he'd sent them, but he never wrote back, according to a screenshot viewed by The Post.
"He would threaten everyone," she said. "He would talk about shooting up schools but no one believed him, no one would think he would do it."
Another 16-year-old said she met Ramos on Yubo in February and that he messaged her asking for her Instagram account. Earlier this month, he reacted to a meme she'd posted that referenced a weapon with a laughing emoji and said, "personally I wouldn't use a AK-47″ but "a better gun": an AR-15-style rifle like the one police have said he used in the shooting, according to a screenshot viewed by The Post.
The Uvalde shooting comes less than two weeks after another gunman killed 10 Black people in a Buffalo grocery store. He live-streamed the attack through the video service Twitch, which removed the stream within a few minutes; copies of it remain online.
The alleged gunman, Payton Gendron, also used the chat platform Discord as a place to save his online writing and pre-attack to-do lists. On the day of the attack, he invited people to his private room, and the 15 who accepted were then able to scroll back through months of his racist screeds and see another view of his attack live-stream. Discord has said the messages were visible only to the suspect until he shared them the day of the attack.
The revelations about the Uvalde gunman's social media activity follow years of complaints from activists and high-profile figures about Instagram's ability to combat its most troubling users. Instagram has said that tackling abusive messages is harder than in comments on public pages, and that it doesn't use its artificial intelligence technology to proactively detect content like hate speech or bullying in the same way.
Instagram users can report direct messages that violate the company's rules against hate speech, bullying and calls to incite violence, and they can block offensive users. But many abusive messages still slip through the cracks. The Center for Countering Digital Hate, an advocacy group, said last month it had analyzed more than 8,000 direct messages sent to five high-profile women and found that Instagram had failed to act on 90% of the abusive messages, despite the posts having been reported.
Facebook's critics have alleged that the ability to tackle dangerous posts could get harder once the company follows suit on its plan to expand end-to-end encryption, which scrambles the contents of a message so that only the sender and receiver can see it, as a default setting on all of its messaging services. Currently, encryption is the default setting on WhatsApp but users only have the option of encrypting their messages on Instagram and Facebook. But the company has argued that as more people flock to private messaging it wants to ensure social media networks are "privacy focused."
In recent years, Instagram has launched new tools to protect teens from predatory users, particularly adults attempting to groom them. Last year, the company began making young teens' accounts private by default once they signed up for Instagram, and they stopped adults from being able to send direct messages to teens that don't follow them. The company also recently announced a "hidden words" feature, which allows users to filter offensive words, phrases and emoji in message requests into a separate inbox.
Yubo said it bans posts that threaten, bully or intimidate other people and uses a mix of software and human moderators to curb inappropriate content. People can block others' accounts or report concerns to a team of "safety specialists," who the company says respond to each person's report.
Researchers have documented that a history of violence or threats toward women is a common trait among gunmen in mass shootings, as evident in the 2016 Orlando nightclub shooting and the 2019 shooting in Dayton, Ohio.
Whitney Phillips, a researcher joining the faculty of the University of Oregon this fall, said social networks could do more to push back on violent harassment toward women, but that the threats on their site are a reflection of a larger "boys will be boys" cultural attitude that normalizes men's bad behavior online and offline.
"When someone says something violent to you or makes some sort of death threat to you, for many women that happens so often that it wouldn't even register with them," Phillips said.
- - -
The Washington Post's Shawn Boburg and Razzan Nakhlawi contributed to this report.