OpenAI has an army of contractors who read real ChatGPT users’ prompts and other data to help improve the chatbot’s responses. The idea is that the contractors provide an, obviously, human touch to OpenAI’s models. Well, not all of the contractors are doing that. 404 Media has found multiple contractors hired to improve OpenAI’s models have been fired for using AI to train the AI. That’s not great for the models themselves, but there is also obviously a great irony in AI training companies working for OpenAI firing people for using AI when OpenAI’s whole thing is to make people use AI at work.

Some AI models already exhibit signs of “model collapse,” which is where AI models further trained on AI-generated text can become worse and worse. In this case, some of the people hired to partially stop that happening are themselves using AI-generated responses to train OpenAI’s models.

One contractor said they see people using AI “all the time and people are let go for it all the time, it’s pretty much the one thing that will get you kicked off ASAP.” The person said, “in a group of thousands there are tons that have been caught.” 

Last week, 404 Media revealed Project Lily, in which OpenAI has hundreds of contractors reading real ChatGPT users’ prompts and conversations which can include personal information. Those contractors then rate and critique the responses ChatGPT generated, including making sure that the responses are not too sycophantic or anthropomorphize ChatGPT. That reporting was based on internal documents and conversations with a person who works on the prompts. 

404 Media has now obtained other internal documents and spoken to three contractors doing work for OpenAI across various projects. Those projects can include more than ten thousand contractors, according to one of the internal documents.

One of those documents says contractors must not use AI themselves for their work. 

“Do not use AI detection tools, or AI yourself,” one document describing the work of contractors who are hired to review the work of other contractors, including catching them for using AI, says. “Do not use GPTZero or any other AI detection tool. They are not reliable. Reviewers may not use AI either, including Grammarly and AI translation, to review, write feedback, or write comments.”

The document continues, “Do not tell evaluators why you suspect AI. It is easier for them to hide if they know what you look for. Judge the overall pattern, not one clue.”

All three of the contractors said reviewers are told not to use AI in their work. Two of the sources said people have been fired or offboarded for using AI. 404 Media granted the contractors anonymity as they weren’t permitted to speak to the press.

The contractors who review other contractors’ work are told to be on the look out for tell-tale signs of AI use. That can include repetitive words, AI-style punctuation — which might include over zealous use of the em dash — and contractors finishing their work very quickly.

In related Slack channels where people ask each other for advice, a lot of people will post an example with the question, ‘Is this AI?,’ one contractor said.

“Usually the answer is yes,” the person said.

One contractor said they used AI while helping to train OpenAI’s models and shared what they presented as their termination letter. It said their employer had identified issues with the “authenticity” of their work. 

“I’m not a bad person or worker. I just needed a little boost and turned to AI to help me which eventually led to my downfall,” the contractor told 404 Media. “I felt no joy in the work or that I was contributing to society in any way.”

Two of the contractors 404 Media spoke to worked for Mercor, an AI-training company that hires the contractors who in turn review ChatGPT-related material. A Mercor spokesperson told 404 Media in a statement: “Our experts are hired for their expertise and judgement, which is essential to the ongoing advancement of AI. Our contracts strictly prohibit the use of LLMs to complete projects and we enforce that. We invest heavily in our tools and systems to detect misuse and ensure our experts comply with project rules and contract terms. When we confirm an expert has used AI to complete a task, we immediately remove them from the project.”

404 Media spoke to a fourth contractor who has worked on training models for various AI companies. They said they sometimes purposefully chose the worst responses because they wanted to actively sabotage the models’ training. 

“I did feel guilty about doing this kind of work at the start,” they said. “I either pay zero attention to the results and choose randomly or purposely choose the [worst] output. I’m not sure  how much of a difference it actually makes since there are hundreds of other people also rating prompt results, but it does feel like I’m getting paid to make AI worse.”

OpenAI declined to comment on its contractors being fired for using AI.

I was trained in giving the DIBELS assessments a few weeks ago along with the other new teachers. The training was run by a consultant from my district’s curriculum company. These assessments mostly involve listening to students read and coding the results to see if their reading is at benchmark. We talked about each assessment, practiced, and poked through the website. One is a comprehension assessment, called MAZE, that wasn’t included in the training. The trainer told us the assessment is online, so we assign it to students, students log in, take it, and that’s it.

I asked if there was a paper and pencil option. The trainer seemed taken aback. Well, there is...but why would you want to do that? I realize I have this reputation now as an anti-technology guy but I’m new and I didn’t want this to be my hill to die on so I let it go and we moved on.

Fast forward: we give the assessments. Afterward I was meeting with the other 5th grade teachers, and we had all observed the same thing: a lot of students guessing on the online MAZE assessment, not taking it very seriously. It wasn’t everyone, but it was enough to be concerning. We agreed that we would give the paper version a shot for our next round of testing.

This is a common story in many schools right now. The details vary, but there are two main parts:

• The assumption that we need to give students lots of tests, and that many of those tests should be digital. Digital is assumed to be necessary and rarely questioned.

• At the same time, many teachers are observing students guessing and giving up more quickly than before, trying to get the test over with as fast as possible.

Mode Effects

Here’s an interesting fact that I think should be more widely known. Lots of tests have gone online in the last few years. During that process, folks who write these tests did lots of experimenting to understand what that transition meant. One of the things they learned is what they call a “mode effect,” which is a fancy way of saying that kids do worse on digital tests.

Imagine the exact same test. One on paper, one digital. On average, students will do slightly worse on the digital version. So test makers correct for this: cut scores are a bit lower on the digital version to compensate. That’s kindof wild to me.

What I’m observing is that students’ relationship with technology is changing. More and more of our students grew up with phones or tablets close at hand much of the time. More and more of our students have been taking these tests for years.

The students who guess aren’t a random sample across the class. They are most likely to be students who are lower-achieving, who find the content hard to begin with. And sure, some of those students will guess on paper as well. But there’s something about screens that reduces the friction to guessing, that brings in the habits students develop playing video games or fiddling with apps, swiping on to the next thing.

The Testing Era

We are in the testing era of education in the United States. Students take state tests at the end of the year in most grades, and then they take more tests to figure out if they’re on track for the state test, and then there are more tests to progress monitor in between those tests. All these tests aren’t neutral. Students are building habits, and for more and more students all the tests seem silly so they build a habit of guessing their way through.

I’ve spent some time digging through the materials and I’m honestly pretty impressed with what the major testing organizations are doing. Here’s TIMSS accounting for mode effects, and PISA comparing student effort between countries — no surprise, student effort has decreased over time. Everyone wants to run lots of regressions to see if digital test results are valid.

Meanwhile, I’m watching students taking test after test, treating tests like a video game to get through as fast as possible. I mentioned student guessing on MAZE but not on the NWEA MAP assessment. MAP has a “rapid guessing” detector: if a student is answering questions too quickly, the test pauses. I check in with the student, remind them to take their time, and restart their test. Students figure this out. And what they learn is not that they should try their best. They learn that they need to guess slowly. Guess. Pause. Stare into space for a minute. Guess again. Repeat.

I don’t think switching to paper and pencil will magically solve all of education’s problems. Teaching will still be hard. Low-performing students won’t magically skyrocket because they took a test on paper. Some students will still guess. Others have built a habit of guessing on computers and they’ll just keep guessing on paper. It will be a bit more work for teachers, though we should get rid of some of these tests while we’re at it to help with workload. My concern is less about the validity of the data. Data is valuable, but teachers today are flooded with data and we will be fine with a bit less. My concern is what we’re training students for. Students should come to school to learn. Students should not come to school to practice gaming tests they don’t care about.

Here’s what I’m imagining: fewer tests, with each test linked to clear outcomes and next steps for teachers based on the data. Tests that we do give are largely given with paper and pencil to improve effort and accuracy. Less pressure on districts to buy and maintain fleets of technology. I have a cart of Chromebooks on the side of my room. They mostly sit there, unused. Which feels like a bit of a waste. But when testing week comes around, it’s tough to imagine making do without a cart for every classroom. I don’t blame my district for investing in all this technology. But seriously, it’s 2026. AI is solving million-dollar math problems. We can figure out efficient ways to assess students with paper and pencil.

After two devoted gray fox parents and their six kits moved into our backyard, we became obsessed with watching them grow, play, learn, and finally, move on

Soon after the introduction of ChatGPT in November 2022, students figured out how to use the chatbot as an answer machine, getting it to do their school work for them. Not surprisingly, that shortcut harmed student learning and led to lower test scores. 

AI enthusiasts sought a solution. The chatbots could easily be instructed to act more like a human tutor, withholding answers, providing hints and guiding students with questions. Khan Academy created one of these “Socratic-method” chatbots, christened it “Khanmigo,” and released it in 2023.

But a new study that came out last month suggests that creating an AI tutor is only one hurdle among many in helping children improve their math skills. Students have to actually want to use it.

Researchers from the University of Toronto followed students in 18 Tennessee middle schools for two years, from 2024 to 2026. Almost all the students tried Khanmigo at first. And many tried to get the AI assistant to spit out the answer. But when Khanmigo refused and instead asked questions or offered hints, students largely stopped using it.

“We observe how students actually used the AI tutor, and the answer is: not much,” Philip Oreopoulos and Nina Low, the researchers, wrote in their draft paper, circulated by the National Bureau of Economic Research in August. “Seeking help with one’s own confusion remained a choice, and most students declined it most of the time.”

The researchers were testing whether Khanmigo could improve the performance of low-achieving students who were at least one grade level behind their peers. These students had an extra remedial math period built into their school day. Some were randomly assigned to use Khan Academy, with Khanmigo available as an AI assistant. The others received their usual remedial math work, which often included online tools such as Waggle, IXL, Zearn and DeltaMath.

The students who received the Khan Academy intervention did better in math than those in the regular remediation classes. But the gains were modest, and no greater than what previous studies had found from Khan Academy practice sessions without an AI assistant.

In other words, Khan Academy appeared to help. But the Khanmigo tutor didn’t add anything. 

Sal Khan, the founder and CEO of Khan Academy, praised the study and wrote a detailed blog post about it, explaining the randomized controlled trial (RCT) just the way he explains a math problem in his videos. In an interview, Khan said the findings confirmed what Khan Academy was seeing internally: Students weren’t engaging much with Khanmigo.

Khan said he doesn’t regret releasing the AI tutor before knowing whether it would be effective. He is confident the early release did “no harm” to students, saying his organization went to great lengths to protect student data, make chatbot conversations safe, prevent cheating and minimize the risk of the bot hallucinating and giving students incorrect instruction. “It’s allowed us to learn and hopefully make the new version even more helpful,” he said.

The version of Khanmigo tested in Tennessee was released in 2023. Since then, Khan said, the company has changed how the AI assistant works. Khanmigo is now integrated directly into Khan Academy rather than sitting in a separate tab that students have to click on. If students ask for help before trying a problem, Khanmigo is deliberately taciturn, offering little more than a hint and a pep talk. But after a student gets a problem wrong, the AI assistant automatically pops up to offer help and work through the problem together with the student. Schools can also choose to turn the AI assistant off completely.

Khan Academy is experimenting with more ways to encourage students to use the AI tutor. Khan said he plans to give students “credit” for using Khanmigo after a mistake, allowing a bot-assisted redo to count toward one of the five correct answers in a row that students need before moving on to the next skill. 

It will be interesting to see whether these changes get students to use Khanmigo more. But use and engagement are only one piece of the puzzle. We still need evidence that when students regularly use an AI tutor, they actually learn more. I’ll be watching. 

Contact staff writer Jill Barshay at 212-678-3595, jillbarshay.35 on Signal, or barshay@hechingerreport.org.

This story about Khanmigo, an AI math tutor, was produced by The Hechinger Report, a nonprofit, independent news organization that covers education. Sign up for Proof Points and other Hechinger newsletters.

The post Students didn’t get answers from Khanmigo. They didn’t want its questions, either appeared first on The Hechinger Report.

Each spring, Patrick Kelly hears the same complaint from teachers in his state of South Carolina: Too many students, particularly in high school, stop doing any schoolwork. 

Students know they can pass their classes anyway because of school policies known as “grade floors.” Schools adopted grade floors — which require teachers to give students minimum grades, commonly a 40 or 50, instead of zeroes — to keep GPAs from being dragged down by a few missed assignments. But in practice, they’ve allowed students to scrape by all year and still end up with a passing grade, said Kelly, director of governmental affairs with the Palmetto State Teachers Association and a high school government and politics teacher. 

“At any high school graduation, I promise you, there will be a child who walks across the stage where every teacher in the room is going to go, ‘Wait, how did they graduate? They didn’t do anything,’” said Kelly.

In May, the South Carolina legislature passed a law banning minimum grades in K-12 schools in an attempt to crack down on grade inflation. The law comes at a time when many people are wondering what an A actually represents. The infiltration of artificial intelligence into school work and lenient grading policies that are a holdover from the pandemic have added to the concern, experts say. South Carolina is now attempting to tackle the problem through legislation, following a similar law passed by Texas in 2009. 

Studies show that although classroom grades and GPAs have been rising for decades, scores on standardized tests have stagnated or fallen. The average high school GPA in 2021 was 3.36, up from 3.17 in 2010, according to a study by ACT, the company behind the eponymous college admissions test. While GPAs were rising, test scores on the ACT actually fell.

Grade floors are just one piece of the grade inflation trend that is “diminishing the value of a high school diploma,” Kelly said. 

It’s a problem that colleges are increasingly wrestling with, too.

Two days after South Carolina’s grade floor law was signed, faculty at Harvard University passed a policy capping the number of students who can receive A’s to 20 percent of an instructor’s class plus four students. The move was a response to perceived grade inflation in the aftermath of the pandemic: A report from the college showed A’s made up two-thirds of the grades given to undergraduates at Harvard in the spring of 2025. 

Some education experts blame grade inflation in K-12 classrooms for contributing to the problem at the college level. 

“It’s hard to hold the line at the higher ed level if you have kids coming into higher ed who are less prepared than they used to be,” said David Griffith, national research director at the Fordham Institute, a conservative-leaning think tank.

Related: It’s easier and easier to get an A in math

This spring, testimony from educators and advocates on grade floor policies alarmed South Carolina legislators. The state is “hurting these kids” by having grade floors in schools, Rep. James Teeple, R-Charleston, said during a committee meeting.

In 2025, South Carolina’s on-time high school graduation rate was nearly 87 percent, a 10-year high point for the state. At the same time, only 33 percent of seniors were considered college ready by graduation — a metric on the state’s report card based on college admissions test scores and advanced classes. More than 1 in 5 students in the state were chronically absent from school — nearly double the rate of chronic absenteeism in the state prior to the pandemic. 

At least 22 districts across the state had official policies preventing students from getting zeros, according to an analysis by Palmetto Promise Institute, a conservative think tank. But other schools also had informal grade floor policies, many of which prevented students from receiving below a 40 or 50, Kelly said. 

One high school teacher in the state, who was granted anonymity because they feared reprisal from their district, said half a dozen seniors passed their class despite not doing any school work this spring because of such a policy. “It’s our fault. It’s not the kids’ fault for figuring out how to beat the system,” the teacher said. 

Giving younger students minimum grades could mislead parents into thinking their children aren’t struggling when they are — studies have shown that parents are more likely to trust their children’s report card grades than test scores. Grade floors are also likely to leave students less prepared for schoolwork as they age, said Dena Crews, president of the South Carolina Education Association, which represents teachers.

“If you’re giving them the minimum of a 60, by the time they’re getting to middle and high school, they’re taking those gaps and deficits with them when they could have gotten the extra help and extra support in the early childhood and elementary grades,” Crews said. 

The new law, which went into effect this school year, threatens districts that don’t comply with the minimum grade ban with the loss of 10 percent of their state funding. It also requires that students turn in any missing assignments before they become eligible for credit and content recovery programs — courses that allow students to make up sections of a class or to pass a class they initially failed. Critics of these programs say they enable students to do no work during class and still pass.

A section of the law also instructs schools not to count most benchmark tests — standardized tests given throughout the year to track student progress — toward a student’s final grade, except under specific circumstances. Eric Duncan, director of P-12 policy for the national nonprofit EdTrust, applauded the legislation for separating benchmark tests from the grading process and encouraging schools to be more transparent with parents. 

“It’s very thoughtful about how important it is for teachers and schools to guide the instruction,” Duncan said. “It’s all part of honesty about where students are and what they need to do to get to grade level.” 

Related: The tricks teachers are trying to fix students’ shortened attention spans 

Teachers in the state have broadly supported the legislation, according to Kelly.

“South Carolina teachers would tell you that policies like grade floors, credit recovery and content recovery — in the way that they’ve been implemented — are leading to students getting diplomas that have no value beyond the paper they’re printed on,” he said. 

Greenville County Schools, the largest district in South Carolina, first instituted its grade floor policy more than a decade ago. It prevented elementary and high school students from getting below a 50 on each of their quarterly report cards. For middle schoolers, the district later banned grades below 50 on all assignments, too. 

Proponents of these policies say the 100-point scale unfairly gives students a bigger opportunity to fail than succeed, since just one zero can depress even a strong student’s average to a failing grade. Grade floors are also meant to prevent “in-school dropouts” — students who do so poorly at the start of the year that they are left mathematically unable to pass even if they aced the rest of the course.

Cinnamon Stetler, a parent of two children in middle and high school in Greenville County Schools, can understand why schools might introduce a grade floor policy for younger students.

“Students should still be allowed to make mistakes and figure things out and develop an understanding of how to do school and how to learn, without it being so much an emphasis on what number you have in a subject,” Stetler said.

But for high schoolers, GPAs matter — to college admissions, for example — and inflated grades have “ripple effects,” Stetler said. And there should be consistency in grading across schools within the same district and state, she said. 

“If you go to a different school where the inherent value of that same mark reflects something different, that’s not a reliable measure,” Stetler said. “It’s problematic for the students; it’s problematic for the parents.”

Susan Stevens, assistant superintendent for school leadership in Greenville, said the district had already planned to eliminate its policy on middle school assignments this year before the state passed its grade floor ban. The district had not planned on getting rid of the grade floor on report cards, Stevens said, though the district will abide by the new law. But she is concerned about other changes in the bill, such as the rule about students completing coursework before they are eligible for content recovery classes. That could result in more students failing altogether and dropping out, she said. 

“We will have to look at how we can support students who are at risk or students who have additional challenges to help them avoid failure,” Stevens said.  

Related: A lot goes on in classrooms from kindergarten to high school. Keep up with our free weekly newsletter on K-12 education.

Grade floor policies started gaining traction in the 2010s alongside a movement known as “equitable grading.” The idea was to remove educator bias from grading and give students chances to recover from low scores. Equitable grading policies, such as grade floors, began to spread more rapidly when schools went remote during the pandemic, as students faced barriers in learning from home, said Griffith of Fordham. 

“The pandemic supercharged it. Frankly, teachers didn’t have much choice. What were they going to do, flunk every kid in America?” said Griffith. “Some level of standards lowering was basically inevitable under the circumstances.”

But even after schools returned to in-person learning, many schools kept the lenient grading policies adopted during the pandemic. A national survey of teachers conducted by RAND and published last fall by the Fordham Institute estimates more than one-quarter of teachers work in a school with a “no zeros” grading policy. A separate study of Washington state’s school policies found that 70 of the state’s 295 school districts had “no zero” policies in 2024-25, up from 30 districts in 2015-16.

“Once you’ve lowered the bar, it’s hard to raise it again,” Griffith said, noting that the use of artificial intelligence by students has made it more difficult to determine the accuracy of class grades.

Minimum grades are just one piece of the larger grade inflation problem. Educators said they also feel pressure from both school leadership and parents to give students grades that are higher than what they earn. Tough graders can expect pushback from everyone, Griffith said, because parents, students and principals are incentivized by higher grades.

There are a lot of incentives for schools to perform well on their state report cards, too, said Derek Lewis, a former school board member of Greenville County Schools and father of a high school student in the district.

“If the number of students failing hurts your school, every school is going to look at how we address that issue, even if maybe the solutions don’t necessarily equate to better learning,” Lewis said. “We focus on what we measure.”

Banning minimum grade policies won’t completely solve the problem of grade inflation, but it gives educators more autonomy to honestly grade students’ work, Kelly said. 

The high school teacher who did not want to be named criticizing their district said they didn’t know a single teacher unsupportive of the law.

“No child wants to fail, but if we keep saving them, why should they try?” the teacher said.

Contact staff writer Ariel Gilreath on Signal at arielgilreath.46 or at gilreath@hechingerreport.org.

This story about elementary and high school grade inflation was produced by The Hechinger Report, a nonprofit, independent news organization focused on inequality and innovation in education. Sign up for the Hechinger newsletter.

The post The emerging fight against grade inflation in K-12 classrooms appeared first on The Hechinger Report.

There are many good reasons to scream. Rent, AI. Your disaster of a love life, getting laid off. Not one, but two of your violins getting stolen, or your parents constantly discussing your hypothetical future husband they invented in their minds and named Raymond, because it sounds like your brother’s wife’s name (Rachel). Or perhaps you just turned 30.

And screaming is more fun with friends! This observation, simple yet powerful, is what inspired me to once again roll up my sleeves and spend over 10 minutes thinking of other stupid things we could do within this theme, then creating flyers to broadcast it.

We obviously needed screaming warm-ups, so as to not strain our screaming muscles. And it’s important to stretch afterwards as well, so we needed screaming cool-downs. We should have a screaming contest to find the best screamer, and they should win a big special prize. And we should have practice screams to ramp up for the one really big one.

I of course anticipated all the questions people would have, and then I wrote different ones that I found funny instead.

It’s been awhile since I hosted an event, over a year has passed since the first (and last) Sit Club (run club without the bad parts), the “Death Duel” (to fix the gender ratio in SF), and Strippers for Charity (self-explanatory).

After all, I’ve been really busy tricking 3 million people into believing in an evil fake polycule, commandeering 1.5% of San Franciscans to do my bidding, starting a youtube channel, running my startup Outernet, moving to a new neighborhood, dealing with a broken refrigerator and fighting Home Depot, an ant infestation, and needing to buy lots of new plants to fill my new home. My autoimmune condition has also been flaring, so I’ve had to cook all my meals from scratch on a specialized diet. But what are these inconveniences but the tension that holds together the woven thread of life? What are these but reasons to scream?1

I didn’t put that much effort into promoting this event, because I didn’t feel like it. I usually just put up some stupid flyers, people take pictures and send them to their friends or post them online, and they go viral.

I often joke I have this duel blessing/curse, the type mischievous, fae-like creatures dole out, where my schemes always go viral, but only from someone else posting them. This time, that person was Greg Dewar, a guy who uses Instagram the old-fashioned way, just casually posting things that spark delight. He snapped a photo and shared it, and that got 18k likes.2 Serendipitously, he shared it on his birthday, and called its virality “a weird birthday present thanks to a bunch of people I don’t know.” So cute! So true also.3

I actually did finally get a shred of Twitter clout from this event, as I shared that when I was hanging up flyers, a man ironically started screaming at me to take them down and following me. And I was like, hey man, maybe you should direct your anger issues toward the sea, not at random strangers (I said this in my head not out loud, as I was too busy moving fast to escape him). I highkey turned lemons (being chased) into lemonade (Twitter clout) with that chasing. Sadly, I have already written about the three times I have been physically chased, and I will unfortunately now have to up that tally to four.

Now my first challenge was identifying a good spot for so many people to scream. I asked my outdoorsy friends for the longitude and latitude of good vantage points. We could do a beach, but a cliff overlooking the sea just felt more powerful. I ventured throughout the Presidio, hunting for a good spot, and discovered4 Vista del Mar. It was perfect: dramatic, castle-like ruins with a gorgeous ocean view.

My second challenge was creating stickers for fun. I had so many ideas that I decided to engage in democracy and let the people decide.

My third challenge was there were so many reporters up my ass to cover this event, and honestly I was like - slow news day, huh? And I told them such. One day all other news will become obsolete in the face of the 𝓇𝒶𝓌 & 𝒻𝑒𝓇𝒶𝓁 media empire anyways, so. Enjoy riding on my coattails while it lasts, buddy.

My fourth challenge was I felt inspired to create a scream zine, but was just so busy and honestly lacked zine-making expertise, so it’d be a big lift to self-learn and then create one. I put out a Call for Artists to two of my incredibly talented friends, Connie and Aileen, and they were both down. I am so, so incredibly grateful to have such talented friends down to clown with my stupid shenanigans, and they absolutely killed it.

Aileen even went rogue and created another mini zine, The Screamer’s Pocket Companion.

My fifth challenge was how to get people to actually show up, and show up on time. Here’s the sitch:

1. There’s severe partifulation (partiful RSVP inflation), where people think an event is funny, so they RSVP yes, then flake or were not even in the state to begin with.

2. People often think my events are jokes, and not actually happening, so I have to prove that they’re serious! Or at least “seriously” happening.

3. Attendees often show up late. You don’t want to start on-time and have half the crowd miss out, but you also don’t want to incentivize people to arrive late by delaying the start.

So to contend with these variables, after scouting out a location:

• I shared the schedule so people knew the event itself would be fun and interesting, and not just a one-note joke of screaming once then going home.

• I sent a pic of the event location to affirm the event was, in fact, not a joke.

• I also set a very narrow event duration of 1.5 hours (so if you arrive an hour late, you’ll miss the bulk of the event) and laid out a specific schedule (so people would know what they’d miss if they arrived late).

• I also broadcasted that the first 100 people to arrive would receive a scream zine and stickers5. This last strategy was extremely effective, many people arrived early, even before me.

Finally, with my five great challenges conquered, it was time to Scream.6

Actually - one more challenge before we get into the event - I was concerned about park rangers showing up and shutting us down. But there was nothing I could do to prevent this, so I knocked on wood and prayed for the best.

Ok back to the main event.7

To ensure no one pulled a scream muscle, we started with scream warm-ups. I had tasked my friend Greta with this, and had no idea what she prepared. And Greta did an incredible job, she actually really got on a power trip with it, which was beautiful to behold. I felt so proud.

Scream warm-ups included fixing your posture, releasing your shame, isolating different parts of the mouth and warming each separately, and performing animal noises. The last one she admitted was just because she wanted to see us do that, not particularly for warming-up.

Greta’s in a tree because she’s just like that. Video credits to Saniya Godil.

After our vocal chords were nice and warm, we held the screaming contest, to find the best screamer. Everyone formed a line, then we drew two lines in the sand, one for the contestants to toe, and one for us to stand behind, to ensure a constant distance for measuring decibel. The scoring was based on volume, duration, and style, and the calculation is proprietary information.

Sam measured sound in decibels, Connie recorded duration, and Sam, Aileen, and I awarded style points. Fun fact: Sam bought a referee outfit to help with the Death Duel to Fix The Ratio, and he’s now used it for four different events (including our friends setting a LEGO world record and a cutting-things-perfectly-in-half competition). Turns out a silly little outfit is often a really valuable investment.

A scream contestant. Video credits to Greg Dewar.

People had real style to their screams, some high pitched, some low pitched, some from the gut, some from the chest, some a long groan, others a short yelp.

Another scream contestant. Video credits to Henna Chou.

Kyana (the runner-up) and a guy whose name I don’t remember! Video credits to Rhea Joseph.

A couple brought their cute lil baby up, who ostensibly screams all the time, but had performance anxiety when it was time to compete. Everyone encouraged the little babe to scream, he opened his mouth, we waited eagerly in anticipation, and a long string of drool came out and slowly dripped to the ground, and the crowd died laughing.

We had to learn the hard way that 39 contenders screaming straight into your face as loud as they can makes your ears hurt a bit.

At long last, I climbed up on a picnic bench and talked people through our practice screams. Starting at just 1% of our power, then ramping up to 10%, then 50%, and finally the one really big one at 100% of our power.

The climactic scream just went on and on, a choir of collective exhalation. Like overlapping waves, each individual’s scream ebbed and flowed, but melded in a cacophony of release, the people clearly needed this a lot.

Video credits to Saniya Godil

Video credits to Sam Sliman.

After our big one, it was time to announce the winner of the screaming contest, Olivia. She performed an encore for the crowd, to teach us what a really great scream sounds like.

Olivia, the #1 Screamer, performing a screaming encore. Video credits to Henna Chou.

Her trophy was a mini microphone labeled “#1 Screamer, a guttural scream at the sea 2026” so that everyone going forth could hear her beautiful scream even louder and more clearly.

I was kinda terrified the whole time that park rangers would show up and shut down the party, considering we did have over 100 people intermittently screaming in public, but luckily it was super chill.8

However, what did show up was a clown on stilts. I thought they might’ve been here for something else, but they leered that they did, in fact, come to scream. More-so I felt they came to help inspire screams in others.

Terrifying! Video credits to Henna Chou.

Once the scheduled screams concluded, screamers mingled, ostensibly using the zine’s icebreakers I had so painstakingly prepared, and ate freshly cut fruit. The crowd would intermittently just be pierced with guttural screaming, a call which others would then join in on.

All in all, it was a great scream. I honestly cannot remember the last time I had such a freeing scream, let alone a community-driven scream, surrounded by my closest friends and people I just met.

All this to say: a guttural scream at the sea healed me.

Appendix:

Here’s the video KQED put together, super cute! And here’s the article from SF Gate.

The Canadian Broadcasting Channel also reached out for some reason, and a bunch of random sites recycled content to patch together their own articles. But the funniest one has got to be the following, which took a picture of my friend Myko from the event and then just AI-generated shit around him. Like, okay.

It was also kinda funny that no one called this event by its formal name, a guttural scream at the sea. They called it a “scream event,” “screamfest,” “scream off,” etc. Like respect its formal title guys.