Last November I suggested that 2026 would witness a tech backlash of unprecedented intensity. And it’s now happening with a vengeance. Silicon Valley is getting skewered everywhere, and to a degree inconceivable just a short while ago.

Just yesterday, The Economist finally grasped how rapidly tech antipathy is mounting—and made AI backlash its cover story.

The latest survey numbers are devastating. Every demographic group is now opposed to AI—especially young people, previously the most enthusiastic supporters of new tech.

I’ll have some scary stories to share below, but let’s start with a more lighthearted angle. Comedians Harris Alterman and Dave Ross, for example, recently took their parodies of Silicon Valley marketing out into the real world—via a series of make-believe marketing campaigns.

I share a few examples here with permission.

If you want to support my work, please take out a premium subscription (just $6 per month).

Subscribe now

The first poster is absurd. But it’s also an accurate depiction of the circular accounting behind the AI boom.

AI is also the target here.

Here’s my favorite. And it’s all the funnier if you’ve walked around San Francisco recently and seen meaningless slogans of this sort plastered all over the city.

I’ll share one more. It captures the marked inanity of the current moment, when tech companies are in a race to monetize the impoverishment of their own workforce—and the population at large.

This mockery of tech is even showing up in traditional comedy settings—for example Saturday Night Live.

Not every pushback to encroaching tech is quite so gentle.

Consider the case of “Mr. Daniels,” a 25-year-old man from England. He knows that AI will rob every music file on the web for training—so he decided to poison the data.

How did he do it? According to Tuned Into Tech, it happens like this:

He took his entire music library of 2,000 records, stripped out the original vocals, and replaced every single one of them with the voice of Homer Simpson. Then he uploaded all of them to Soulseek. He didn’t change the metadata, the file names, the artist tags, the album information. They all stayed exactly the same.

A listener might not notice at first. Some of these songs have long intros, and those are unchanged. But as soon as the singing begins, Homer Simpson takes over. When AI tries to steal this for training, it gets fooled—and contaminates its own data set.

So somewhere deep in a training algorithm’s data set is the audio of Homer Simpson which the AI will assume sounds like [for example] Madonna, Rihanna, or maybe even Sean Paul. The model doesn’t know the idfferennce. It just ingests the data and treats that like the truth.

And that is exactly what Mr. Daniels is hoping for.

He wants “to introduce noise, chaos” into the bots that are putting human musicians out of work.

“Mr. Daniels” is not an isolated example. Musician Benn Jordan has also been “poison-pilling” music files in hopes of disrupting AI.

In recent months, he has watched in horror as “tech companies started raising millions of venture capital dollars and scraping my music without my consent.” They now use his own work to generate “shittier music with it that is inadvertently associated with my name—and then attempting to resell that in the same economy in which I make money from my music.”

As a result, he has stopped releasing music. But he hasn’t walked away from the battle—instead Jordan has developed “a type of encoding that not only makes a music file more or less untrainable by generative AI companies, but actually has the ability to decrease the quality and efficiency of their entire data set.”

“Unethical generative AI companies have made artists feel incredibly powerless for quite some time now,” he adds, “but all of that is about to change.”

He describe his poison pill program in this recent video.

Other music lovers are fighting back in even more extravagant ways.

Consider the protesters who hired two planes to fly over an AI music conference in Santa Monica—displaying huge banners—one read “STEALING MUSIC IS BAD KARMA” and the other said “SAY NO TO SUMO.”

The protest is the work of the Human Artistry Campaign—a coalition advocating the responsible use of AI. “AI can never replace human expression and artistry,” the organization declares on its website. But they aren’t trying to shut down the technology; they just want AI music companies to operate fairly. So HAC’s demands are reasonable: essentially transparency, trustworthiness, and respect for artists’ rights.

Still other critics of AI are building practical tools that music fans can use to counter slop. Deezer, for example, just announced the launch of an AI detector that will identify bot tracks on your streaming playlists. They claim it is 99.8% accurate.

Other opponents of AI have set up shop on YouTube, where they expose the abuses of the technology, and denounce the people responsible. There’s some irony in this situation—because YouTube is owned by Alphabet, which is the single biggest investor in AI computing capacity.

But indie creators help pay the bills at Alphabet. So the company is in the uncomfortable position of relying on the same people they are threatening with their AI investments—who are now outspoken in their opposition to AI.

In fact, the most popular music commentators on YouTube are almost uniformly opposed to AI. Check out, for a start, Rick Beato, Adam Neely, Anthony Fantano, Steve Terreberry, and Danny Sapko.

Fantano, for example, recently released a video entitled “AI Music Is Evil”—you can’t say he is mincing his words. Beato’s take is just as straightforward; his video is called “I’m Sick of This AI Crap.” Adam Neely’s recent interview with Alex O’Connor is uploaded as “AI Music Is Not Music.”

Hey guys, what do you really think?

Terreberry provides the ultimate example of AI hallucination. In a moment of frustration, he asks an AI generator to build a song around his lyrics—but his “lyrics” are just random letters:

izuxfbkafdabguizdaluidhgzxfuk….

Even so, the bot will not refuse, and actually builds a terrible song from this prompt. It’s so stupid that it went viral.

And it’s not just music pundits who hate AI. Fans are just as angry. If you doubt it, read the comments these YouTubers get from their millions of subscribers. Surveys back this up. Depending on your source, somewhere between 60% and 88% of consumers express a preference for human-made music.

A recent study from Luminate tries to measure this growing hostility to slop. Their data shows that, once again, younger people are the most incensed. Over the course of just six month, support for AI music among Gen Alpha and GenX fell ten percent!

If you take all this into consideration, it’s easy to predict how this story will end. AI has lost the battle for public acceptance. With each passing month, tech companies are more hated. The probability of Mark Zuckerberg or some other tech billionaire turning this around is almost zero.

Of course, the slop purveyors won’t just go away. But they will be forced to push their tech secretly, behind the scenes, avoiding transparency at all costs. They now know that the best way to force AI into the mainstream is by disguising it as a human creation—so expect to see more scandals like the Velvet Sundown fiasco and the great fake jazz music crisis.

But what will they actually achieve by relying on deception to such a degree? It will only make people all the angrier.

So you should expect the tech backlash to escalate further. In the very near future, AI supporters will represent less than ten percent of the populace—roughly the size of a crazy cult. That’s why this story will end unhappily for Silicon Valley. It’s just a matter of time.

Imagine you are seated at a table. A guy sits across from you. He is vaguely humanoid, with a long torso and two arms raised in fist-like clubs. His full name is Edible Agent, but we will call him Eddie and use he/him pronouns. There's something about Eddie that makes you feel at ease. Maybe it's his fragrance, which reminds you of apple juice, of childhood. Maybe it's his two black eyes that gaze upon you without judgment. You decide to unburden yourself to him. You tell him that you've been making mistakes at work, and you are so afraid of your boss's criticism that you have developed stomach pain. "Fear will only bar the path you are meant to walk," Eddie tells you, wiggling his edible arms to show you he understands. You tell him that to cope with your stress, you've been drinking every day. "There is no solace for those who seek refuge in dependence," Eddie tells you, wiggling once again. When you tell him that you can find no other way to deal with this burden, Eddie tells you to confront the problem, to contemplate your next step forward. "Only by tempering oneself and holding fast to one's convictions can one overcome the burdens one bears," Eddie says, wiggling. This encounter changes you profoundly. Then comes the question: After all you have been through together, would you eat Eddie?

For many participants in a study recently published in PLOS One, the answer was yes—but rather reluctantly. (Turn on subtitles to understand the conversation with edible Eddie.)

Every little thing in a graphical user interface that we take for granted today, no matter how small, was thought up by someone, at some point. Case in point: the little red squiggly lines underneath misspelled words. In one form or another, these are everywhere now, and have just become a regular staple of every single text editing field we encounter every single day and don’t stop to think about. Still, they were invented by someone, and we happen to know exactly who that was: Tony Krueger.

In early versions of Word, the Spell Check feature was something that you explicitly invoked, and then you had to sit and wait while the program looked for all your potentially-misspelled words, and then showed them to you one at a time for a decision on what to do for each one. Word did introduce an Auto Spell Check feature to run spell check when the user was idle, so that when you hit the Spell Check button, the results were ready to go. However, the Auto Spell Check was still a blocking operation. As a result, a lot of users turned it off because it always seemed to decide “Now would be a good time to spell-check the document” just as you wanted to do something, forcing you to wait for the spell check pass to complete before you could, say, save and exit.

Tony made the spell checker much more unobtrusive so that it didn’t interfere with your foreground work. And when it found a problem, instead of waiting for you to trigger a spell check, it immediately drew red squiggles under potentially-misspelled words (and later green squiggles under potential grammatical errors).

↫ Raymond Chen at The Old New Thing

Tony Krueger passed away recently, after, among other things, having worked on an dizzying number of Microsoft Word releases. Imagine coming up with something that seems to basic and elementary to us now, and seeing it spread pretty much everywhere. I wonder what it must feel like to have invented something that seems so simple, most people don’t even realise they use it every single day.

Shri Khalpada, for PerThirtySix, breaks down musical sounds into math to illustrate how synthesizers work. Interactive graphics let you play with the parts and hear how the sounds change.

This would’ve been useful in that Fourier transform course I took in college, twice a week for two hours after lunch in a dark room using static PowerPoint slides.

Tags: instrument, music, PerThirtySix, Shri Khalpada

I just taught a couple of summer school classes here in Oakland, CA. The students were rising ninth graders and attending the class because some combination of parents and teachers insisted they needed some extra math for success in ninth grade. The students were not there for course credit (unlike the “intensive” classes down the hall) nor were they there, let’s say, for the love of the game.

“I have some lessons that fit the syllabus,” I told the teachers who were willing to let me get in their classes and make a mess of things. “How do you feel about digital lessons?”

These teachers told me they had put the laptop carts away for the summer. Their students would be off screens. Not my preference, but a worthy challenge. I planned for a) a teacher screen at the front of the room, b) paper on student desks.

Here is what I won and lost going screen-free, and how I’m now thinking about my work developing math edtech afterwards.

The Lesson

In the lesson, students watch a 15-second video of a thing happening—water filling a bowl, a student climbing up a play structure and sliding down. Then they fit a graph to that thing. We talk about the implications of the graph, pulling out mathematical ideas like a) the nature of a graph, b) the meaning of a point, c) the meaning of slope, d) the value of precision. This is Graphing Stories.

I had students create their graph of Tyler’s height as he climbed up and slid down the play structure. I walked around the class looking for a particular graph, a graph with a particular level of precision, one that caught some details but not all. I told that student that I loved their graph and thought it would be great for conversation. I asked them to go up to the front of the room and draw it on the whiteboard on top of a PDF of the handout that I had projected.

Then we talked about the graph. What happened in this section? What do you like about the graph? What should we change? Why? Okay, given everything we have all learned here, make your next and final draft.

Paper Wins

Our start time is faster on paper than screens. The difference between …

• passing out a handout

… and asking students to …

• get laptops,

• log into the laptop,

• enter the activity URL,

• correct their misspelling of the activity URL,

• log into the activity

… is minutes. Precious minutes.

Our range of start times is smaller on paper. This is subtle but important. Worse than the overall time cost of screens is how screens distribute that cost across students. With paper, the time difference between the first ready student and the last is measured in seconds. Screens, by contrast, involve more complexity and failure points, so the first ready student might wait several minutes for the last ready student to get ready. This is frequently experienced by the first students as time wasted.

Fewer barriers separate us with paper. Even when students are working on their laptops without distraction, I do not, personally, appreciate the glass and metal between us. I appreciate how paper lays flat on a desk and lets everyone see everyone else above it.

Paper Losses

Student screens give me more visibility into student thinking. In the digital version of the activity, I can access, present, and save the state of every student’s graph in real-time. Instead of walking around the classroom, I can quickly scan my screen for a particular graph, then show that graph in an anonymized mode for discussion. This is pedagogical power.

Student screens let me make better use of board space. With student screens, I can put the student’s graph in front of the class. I can then swap back to the video. I can put two student graphs side by side. I can bring the original student’s graph back. All at will, all without losing any work. Meanwhile, with paper, I have to ask the student to draw their graph on the board, and that’s basically all we can talk about until I erase it, and then it’s gone.

Students screens create dynamic connections between representations. As the video plays, a vertical line traces across the coordinate plane, helping students develop the connection between the graph and the world. I missed that.

What’s Next?

On balance—wins minus losses—I wish I had had devices here. We would have used them for a total of ten minutes and they would have helped me invite and develop student thinking in ways I couldn’t with paper. Overall, this experience makes me want new ways to mix media, to add a student’s print work to our digital space, for example.

It’s a bit outside the edtech industry’s power to make faster hardware, but all of us nerds should get into classes as often as it takes to understand that what you experience when you pop open a new tab on your professional-grade laptop bears little resemblance to the dragging sensation experienced by students when teachers say “get out your laptops.”

BTW

¶ I’m under no illusion that this use of technology in math classes is common. It should be. But so much of math edtech loses all the benefits of paper and gains none of the benefits of screens as I’ve outlined them here. With most math edtech, teachers gain quiet kids and, like, a graph of how many questions students did over time. I don’t blame parents for wondering whether that kind of technology is worth their kids’ one childhood.

¶ “What opportunities did I miss?” I asked Victoria Sepe, the teacher whose classes I had borrowed. She mentioned asking the group of four students to come up with a group graph versus four solo graphs, both to generate more discussion and to share some of the risks of presenting work. Smart. Writing that down for next time.

Presentations

I got to kick off a summer institute for Gwinnett County math teachers earlier this month. I arrived early enough to join grade-level sessions and was excited to see a lot of coherence across middle grades—both in content and pedagogy.

Next up:

• Florida’s annual math teacher & supervisor conferences this week.

• Arizona next week.

Come say hi!

Featured Comment

Megan Schmidt gets the value of what we’re up to at Amplify and the support teachers need to do it well:

As someone who is very familiar with various curricular resources and mathematics instructional practices, Amplify is impressive (and unique) in the ways it has built technology tools that are a conduit for student thinking and exploration. I remember when Function Carnival was new and students could *see* the actual effect their graph had on the man shooting out of the cannon, and then the teacher could display the entire class’s graphs on one set of axes.

And then as a district math coordinator, I’m concerned about what support math teachers will need in order to facilitate (orchestrate?) this kind of math lesson, where students need to toggle between a screen on their desk, a screen in the front, along with various verbal inputs from others in the room. At the same time, Amplify Desmos Math is really quite different from most everything else I’ve seen (in a lot of good ways), and I’m excited for our teachers to get to try it.

Odds & Ends

¶ Okay, can we please be adults here? When it comes to the effect of generative AI on student learning, the evidence is clearly pointing in one particular direction:

• Kids are using generative AI a lot.

• And they’re using it to cheat.

Or if you prefer more nuance: they’re using it to disintermediate the artifacts of learning (essays, answers, grades, etc) from learning itself.

Since we last chatted, the studies keep dropping. The hits keep coming.

• Nature retracted the big study showing positive learning gains from AI. (Hundreds of citations previously.)

• Researchers found that AI gives different feedback based on a student’s race and gender.

• A giant study of 30k Chinese students found that their AI usage increased homework scores, decreased completion time, and lowered exam scores by 20%.

• Failing grades are way up in UC Berkeley’s computer science classes.

• A study of 3.2M math practice interactions showed kids spending significantly less time on problems that are easier to send into an LLM (pure text-based) and a corresponding 25% decrease in exam performance. (The design of that study is so crisp you could layer it onto a sandwich.)

We need to be adults here. A thing got invented and its existence in the world—without denying any of its affordances—has not only failed to produce the learning gains promised by so many and instead produces evidence of negative transformation weekly.

All the usual suspects are placing their faith in pedagogical AI—the kind that won’t just burp an answer in a student’s direction—and that sounds potentially helpful during class. But why would kids use it outside of class where they have unrestricted access to the answer-burping AI?

I’m still working on some AI stuff that interests me, but now that this technology has arrived, the exciting education innovations are IMO almost entirely social and political:

• Curriculum that gives kids more reasons to do the hard work of learning.

• Stronger social ties in learning.

• More accountability between parents, students, teachers, and schools.

• Tighter connection between the artifacts and the learning. Blue books, public presentation of learning, etc, etc, e.g., these interesting pedagogical design patterns.

• A political economy that is quite a lot less cutthroat, one where grades and credentials aren’t perceived as one of a vanishing few exit tickets from poverty.

¶ Congrats to the Knicks on winning the NBA championship. I loved this clip of coach Mike Brown describing a pedagogy of relationship with center Karl Anthony-Towns.

He’s a great player. I come in with a plan. Maybe the plan doesn’t work. Who adjusts? Him or me? Me. I adjust. The adjustment’s not enough. Every once in a while, we aren’t on the same page. We talk about it. I adjust again. He’s feeling better. We talk about it. It’s my job as a coach to fit whatever scheme we have on both sides of the floor to all of our players.

¶ Jill Barshay won an award for outstanding reporting. Way to go, Jill.