Welcome to Slopworld

Summary

Paul and Rich tackle three AI topics in rapid succession: labor, security, and slop. First, they examine a Yale Budget Lab report on AI’s actual impact on the labor market, which reveals a surprising finding — despite all the hype, only the engineering/tech industry is seeing significant AI adoption above expectations. The rest of the economy, from forestry to legal to management, is barely moving the needle. Rich offers a profound explanation rooted in organizational change management: people who become experts at particular ways of working resist change not because they fear job loss, but because their expertise has calcified into organizational processes and systems that actively resist transformation.

Second, they discuss the fascinating case of Daniel Stenberg, creator of curl (the foundational internet library), who has been besieged by AI-generated bug bounty reports that are mostly garbage. But then a plot twist: Joshua Rogers used AI tools intelligently to find 22 legitimate security fixes in curl’s heavily-vetted codebase. The lesson is clear — expert-guided AI is a superpower, while naive AI-assisted bug hunting is just spam. Finally, they confront “slopworld” — the flood of AI-generated video from tools like Sora and Meta’s Vibes. Their verdict: when you have unlimited inventory, nothing is interesting. The word “slop” perfectly captures forgettable, strip-mall-grade content that commoditizes everything.

Highlights

”If you have unlimited inventory, nothing is interesting”

yt-dlp --download-sections "*26:10-27:10" "https://www.youtube.com/watch?v=kMz317CHnIk" --force-keyframes-at-cuts --merge-output-format mp4 -o "kMz317CHnIk-26m10s.mp4"

“If you have unlimited inventory, nothing is interesting. If you showed me one of these outside the AI slop revolution, I’d be like, ‘Oh, wow. That’s interesting.’ Because I’m not seeing it a thousand different ways all day.” — Rich Zotti, 26:10

”Only engineers are above expectations in AI adoption”

yt-dlp --download-sections "*11:20-12:20" "https://www.youtube.com/watch?v=kMz317CHnIk" --force-keyframes-at-cuts --merge-output-format mp4 -o "kMz317CHnIk-11m20s.mp4"

“There is no revolution happening except in engineering. When you work in tech it feels like the world. But man, the fisheries industry — not really all over AI.” — Paul Ford, 11:20

”The body rejects the transplant”

yt-dlp --download-sections "*15:10-16:10" "https://www.youtube.com/watch?v=kMz317CHnIk" --force-keyframes-at-cuts --merge-output-format mp4 -o "kMz317CHnIk-15m10s.mp4"

“It’s sort of like you get the liver transplant and then you wait to see if the body rejects it or not. This is a transplant of new technology into the culture. And very often the body is like, I just don’t want any part of this. I’d really actually rather not survive.” — Rich Zotti, 15:10

”The robots intelligently guided find dozens of credible bugs”

yt-dlp --download-sections "*20:40-21:40" "https://www.youtube.com/watch?v=kMz317CHnIk" --force-keyframes-at-cuts --merge-output-format mp4 -o "kMz317CHnIk-20m40s.mp4"

“I’ve been programming for 25 years. I’m not going to find a bug in curl. But the robots intelligently guided find dozens of some of them pretty serious, credible bugs to harden and improve the product.” — Paul Ford, 20:40

”Slop is forgettable strip mall garbage”

yt-dlp --download-sections "*30:40-31:40" "https://www.youtube.com/watch?v=kMz317CHnIk" --force-keyframes-at-cuts --merge-output-format mp4 -o "kMz317CHnIk-30m40s.mp4"

“Slop. It’s a perfect word. Forgettable strip mall garbage. It’s the visual equivalent of auto-generated filler music for malls.” — Paul Ford & Rich Zotti, 30:40

Key Points

• Yale Budget Lab Report (7:30) - “Evaluating the Impact of AI on the Labor Market” published October 1, 2025, shows AI adoption is far slower than expected
• Only Engineers Above Expectations (11:20) - Engineering is the only field where AI adoption significantly exceeds expectations; other sectors barely register
• Tech Industry Bias (11:40) - Our industry bias makes AI seem bigger than it actually is in the broader economy
• Yankees Baseball Opener (0:00) - Rich’s Yankees are down 0-2 to the Blue Jays in the ALDS; he compares baseball’s uncertainty to starting businesses
• Organizational Calcification (13:20) - Rich explains how organizations resist AI not consciously but because processes and systems have calcified around existing expertise
• Liver Transplant Metaphor (15:10) - AI adoption in organizations is like a liver transplant — the body may simply reject it
• Risk vs. Growth Mindset (16:00) - Large organizations have risk-based mindsets, not growth-based ones, making AI adoption harder
• Curl Bug Report Spam (17:30) - Daniel Stenberg, creator of curl, has been overwhelmed by AI-generated fake bug bounty reports
• Joshua Rogers’ AI Success (19:30) - Used AI tools to find 22 legitimate bug fixes in curl’s heavily-vetted codebase
• Expert in the Loop (21:30) - The future is expert-guided AI, not replacing experts; go be an expert is the message
• Parallelization (21:00) - Simon Willison is writing about using multiple agents at a time for exponential engineering power
• Sora and Vibes (24:30) - OpenAI’s Sora and Meta’s Vibes generate 5-second goofy videos that are essentially slop
• Unlimited Inventory Problem (26:10) - When you can generate anything, nothing is interesting — it all gets commoditized
• James Blake’s “Like the End” (28:20) - Example of treating AI video as a creative tool rather than final production
• Constraints Drive Creativity (27:30) - Professional music production lesson: limit your choices and go really slow
• Slop as a Perfect Word (30:40) - The term perfectly captures forgettable, strip-mall-grade AI-generated content

Mentions

Companies

• Yale Budget Lab (7:30) - Published the AI labor market impact report
• OpenAI (24:30) - Makes Sora video generation tool
• Meta/Facebook (24:30) - Makes Vibes AI video tool
• Aboard (3:40) - Hosts’ company; could build a “manager finder for the Yankees”
• Toronto Blue Jays (0:30) - Beating the Yankees in the ALDS
• Sony (0:00) - Made AIBO robot dog 25 years ago

Products & Technologies

• curl (17:00) - Foundational internet library whose security profile is critical to the internet
• Sora (24:30) - OpenAI’s AI video generation tool
• Vibes (24:30) - Meta’s AI video generation tool for phones

People

• Paul Ford (0:00) - Co-host
• Rich Zotti (0:02) - Co-host, Yankees fan
• Daniel Stenberg (17:00) - Creator of curl, besieged by AI bug report spam
• Joshua Rogers (19:30) - Found 22 legitimate bugs in curl using AI-assisted tools
• Simon Willison (21:00) - Writing about parallelization with multiple AI agents
• James Blake (28:20) - Indie electronic artist who made “Like the End” using AI-generated video creatively
• Aaron Boone (0:10) - Yankees manager who Rich wants fired
• Martha Gimble, Molly Kimble, Joshua Kendall, Mattie Lee (8:00) - Authors of the Yale Budget Lab report
• Cardi B (24:00) - Paul likes her vibe; first album was really good

Surprising Quotes

“I consider myself actually a transformational expert. I am AI. No, but seriously, you’re not good news.” — Rich Zotti & Paul Ford, 13:00

“I’ve been programming for 25 years. I’m not going to find a bug in curl. But the robots intelligently guided find dozens of pretty serious, credible bugs.” — Paul Ford, 20:40

“The reason I love baseball is the best teams with the most expensive payrolls can only win like 65% of their games. You can’t dominate in baseball. I think about baseball when I think about starting businesses.” — Rich Zotti, 1:20

“It’s the visual equivalent of auto-generated filler music for malls. That’s all it is. It’s just this forgettable stuff that you’re just going to hear and then you’ll want Panda Express.” — Rich Zotti, 30:50

Transcript

0:00 Hi, I’m Paul Ford. The Yankees crapped the bed. We’ll come back to that. What’s your name? I’m Rich Zotti. And this is the Aboard Podcast. It’s the podcast about how the Yankees crapped the bed to the Blue Jays, but also it’s about AI and software. I don’t care much about baseball, but I know how bad that was. Play that theme song.

0:35 All right, Rich. So, if you had one chance — look, what software do we need to build for the Yankees? I would build how-to-fire-a-manager software, like a manager CRM to go find a new one. I’m not an Aaron Boone fan. He’s the manager of the Yankees. We could actually do that at Aboard. If you go to Aboard and make a manager finder for the Yankees, go share that link back with us.

1:10 So check us out at aboard.com. We can probably save baseball with our amazing enterprise software. The reason I love baseball is the best teams with the most expensive payrolls can only win like 65% of their games. You can’t dominate in baseball. I think about baseball when I think about starting businesses. You oscillate between absolute uncertainty and other days you feel like you’re gonna dominate the world. But the Yankees suck right now.

1:50 The other thing is that it’s incredibly boring and everybody complains about it. I’m kind of used to it. It’s trying to power through the AI generated slop that everyone is consuming. I’m talking about business though, not baseball. They’re both boring and you might not succeed anymore.

3:30 How was your weekend? I was back on the bike. I go out to the Rockaways the long way out through Shirley Chisholm Park. On the weekends I want a good steady ride. So out I go to the Rockaways and then I pull a little trick. I jump on the ferry, take it all the way around. You see Coney Island and Seagate. You stop in Brooklyn and Wall Street and then you ride your bike over the Brooklyn Bridge and go home.

5:00 And then Sunday was garage sales. The social event of the calendar in my crunchy Brooklyn neighborhood. We made $670 for charity. We’re going to use that to feed people in Flatbush. I sold all the old prints from our last company that had been on the walls. People want stuff for their walls. A teacher wanted one with pictures of books for a classroom.

6:20 So that’s our lives. Now, we’re going to hit three things fast in this podcast. Labor, security, and slop. These sound interrelated. Not really. It’s all through the lens of AI.

7:30 Here is from the Budget Lab associated with Yale. “Evaluating the Impact of AI on the Labor Market: Current State of Affairs” by Martha Gimble, Molly Kimble, Joshua Kendall, and Mattie Lee. October 1st, 2025. So if you are out in the world reading about AI and what it’s going to do to jobs — the robots are going to take over jobs and people will be out of work. That’s the most reductive sentiment.

8:30 According to this report — go to figure 11 and figure 22, they’re the hot ones. What this report keeps doing is going like, okay we looked at it and it seems a little early to be drawing all these conclusions. Figure 11 is this almost flat line showing the changing proportion of workers in occupations exposed to AI.

9:10 You and I have an AI company and we use this stuff all day long and we’re technologists. We’re like, “This is the big change coming to our industry.” We’re in a tiny tiny bubble. When you look at the actual broader economy, aside from the technology industry and the media industry, it seems much much much bigger than it really is.

9:50 Go down to figure 22. That’s the gap between actual and expected AI usage by major occupation. Nobody’s really expecting forestry people to be using a lot of AI and guess what? They’re not. Office and admins not expecting too much, but they’re actually using it quite a bit. But really everybody’s kind of within the range of expectations with one exception — engineers. Us.

10:30 Our bias, our toy. When you work in tech it feels like the world. But man, the industries where it really matters: sales, office and admin support, management, engineering, and a little bit of CFO stuff. There is no revolution happening except in engineering. It’s very early.

11:20 How long to really bake AI into culture everyday like the internet has been? I consider myself actually a transformational expert. I’ve gone into large organizations where leadership has asked for significant change. I am AI. No, but seriously, you’re not good news. Even if you’re not going in to lay people off, they weren’t able to make the change themselves so they had to call somebody else in.

12:00 I want to make three points. First, people who become experts in particular ways of working are very resistant to change, not because they consciously worry about their jobs. They are resistant because they’ve gotten incredibly good at it. The gravitational pull of doing it that way is incredibly strong.

12:40 Second, if you take all those atomic units of expertise and put them together, it embeds itself in the culture and how they work. Systems and processes are in place around those atomic units. The organization is literally optimized to work a certain way. You’ve introduced transformational change that the org subconsciously looks like it’s resisting. It’s not resisting — literally everything is calcified.

13:20 The processes have calcified, the systems have calcified. There are literally systems that exist because of the way people work that have to get uprooted for this change to be transformational. How long does it take for those systems to be uprooted? I’m not even going to guess.

13:50 Third point — I would be wary of stats like these because people are still pretty embarrassed and ashamed about using AI a lot. They factored that in though. I think we really are seeing slower change than we thought. Old habits not only die hard, they thrive and are defended. Both at a personal and organizational level it takes a lot to uproot them.

14:30 I use a metaphor for this — it’s like you get the liver transplant and then you wait to see if the body rejects it or not. This is a transplant of new technology into the culture. And very often the body is like, I just don’t want any part of this. I’d really actually rather not survive. It is hard to get people off a spreadsheet, let alone embed and integrate this whole new way of working.

15:20 Large orgs and successful people in large orgs have a risk-based mindset, not a growth-based mindset. That’s so different from startups. Tech world comes in and is like, don’t you want growth? And the CEO is drooling because that summer house is screaming to him. So he’s like, let’s get this tech growth jammed into our carpeting company and everybody on the floor is like, if you screw this up I won’t be able to get you the margins with the suppliers. Resist. People resist.

16:20 Change is gradual. Number two. Daniel Stenberg — first of all, there is a low-level piece of technology called curl. Curl is a command line tool and programming library that lets you get data from the internet. Things that talk to the internet on your phone or computer often use curl. It is infrastructure for internet stuff.

17:00 Its security profile is very very important because a vulnerability in curl is a vulnerability in the internet. And like all open source projects, it’s really just a few people and underfunded. The creator has been besieged by AI-generated bug reports seeking bug bounties. He’ll be like, “That’s a library we don’t even use. What are you doing here?” And then they’re like, “Oh, sorry.”

17:40 It becomes really clear. He’ll reply and then obviously LLM — you know, “Thanks. That’s a great point.” Because there’s bug bounties on the other side. If you find a security bug, you can get a little money for it. They’re spamming him with very technical looking spam.

18:10 He’s kind of become somebody who in public is like, “This is a mess. We have to stop taking these reports. I can’t take care of bot output so that you can make your money.”

18:40 But then this guy Joshua Rogers sent a massive list of potential issues in curl that he found using a set of AI-assisted tools. This is not just hacking with ChatGPT. This is somebody who’s like, I’m going to use AI to analyze this code. The inflection point is: man, they’re spamming these guys, but if you use it right, you can solve problems.

19:20 The guy just put 22 bug fixes in with security issues and other issues. He used AI wisely to interrogate the codebase, find holes, plug them up. The response to all that spam is if you use it right, you can solve problems. This is one of the most vetted pieces of code on the internet.

20:00 I’ve been programming for 25 years. I’m not going to find a bug in curl. But the robots intelligently guided find dozens of pretty serious, credible bugs to harden and improve the product. Those are reviewed by humans and brought in. The robots were unleashed by very knowledgeable, thoughtful engineers that knew what they were doing.

20:40 Simon Willison is starting to write about parallelization — using multiple agents at a time. This is when you start to get exponential weirdness in engineering. Computing is one of those things where you can turn that labor knob up because the better it gets at it, you can do more of it at the same time.

21:10 Good, thoughtful, senior level expertise, architect level expertise use these tools much more effectively. The tools are a trap if you think you can shortcut. This is a message to the youth of the world — these tools are putting forward so much magic and convenience in their promise. But you’re going to still have to put your head down and understand pretty tricky, difficult comp sci concepts. Go get that masters in comp sci because AI is not going to go there just yet.

22:00 Expert in the loop is the future. That’s different from becoming a junior engineer and kind of trudging through. It means you want to apprentice to an expert as soon as possible. Beware the shortcut — it’s not real. The expert will have a lot of opportunity because they become superheroes with these tools in hand.

22:40 One side — a lot of people kind of hacking around making a mess. The other side — people using this stuff to clean something up that everybody thought was already pretty clean. I need a palate cleanser. This was a little nerdy.

23:10 All right. This was the week of slop. Like slop just started showing up everywhere. So this is Facebook Vibes and Sora from OpenAI. What’s Vibes? It uses AI to generate 5-second goofy videos for your phone. Type some stuff in and it’ll generate a thing. Groundhog flying like Superman. That kind of thing.

23:50 I gotta say like it’s just slop. I’m looking at the Sora ones right now and it’s just pouring coffee and there’s bugs, pixelated bugs and jellyfish. This is their best. A guinea pig at the swimming pool. These prompts, this sort of — it means nothing. It doesn’t even look that cool.

24:20 Apparently the new version of Sora is really good at putting your face into things so you can do stuff with your friends, which I think will be okay for the group chat. We stopped using any AI images for the most part in our marketing because they look so bad. Glazy, glossy and weird. It just looks like you took the easy way out.

25:00 I will still occasionally create an image and drop it in Slack to punctuate a point and everybody rolls their eyes. I think that’s the future for creative work with this stuff. It just doesn’t seem to be getting to that meaningful level.

25:30 Here’s my take. If you have unlimited inventory, nothing is interesting. If you showed me one of these outside the AI slop revolution, I’d be like, “Oh, wow. That’s interesting.” Because I’m not seeing it a thousand different ways all day. It’s a numbness that is kind of taking hold. Same with social media. You got to constantly reach for the stimuli hit.

26:10 My problem isn’t the output. It’s kind of crazy that these videos are being produced so quickly. But if you give me an endless supply, then you’ve commoditized everything and nothing can be interesting because my brain can only ingest so much.

26:40 As a hobby, I keep trying to learn music production. And over and over every lesson from every professional, there seem to be two things: limit your choices and when you practice, go really slow. That’s it. There’s kind of no other principle that matters. Limit your choices. Go for the first best take possible, record it, move on.

27:10 We’re ending up in this world of infinite possibility, infinite visual style, infinite characters, hippos running around dancing. It’s a trend. It’ll wash away. When the constraint system shows up and people can start to understand how you’re creating within the constraint system — for a while it was exciting because people were learning how to prompt. But now we kind of all know that. There’s no boundaries to push against.

27:50 James Blake came out with a video recently where he used AI. It’s called “Like the End,” which is just clips of AI video. And I liked it. He treated it as a creative tool rather than as final production. He had a message to share about the world. He wrote a song, has nothing to do with AI, and cobbled together visuals that are almost a commentary about us and generative content. Dark and dystopian and kind of fun.

28:30 He put a little bit of himself into it rather than wrote a sentence and said, “I’m going to put this on YouTube along with my music.” Putting some effort and energy and a little bit of yourself into a thing. We’re not going to create creative works with a sentence. If you can, then everyone will, and then nothing is interesting.

29:00 We want constraints and we want human intelligence and rarity. We want human individuality to inject itself into the tools we use. Photoshop caused a big hubbub. Unions protested synthesizers. People are always going to see these things as threats. It always reverts back to what did you bring to it? How did you use it as a creative tool rather than automated output?

29:40 So you’re not inherently anti-slop? I’m not interested in it. Do I want it to go away? Who cares. Everybody’s drooling looking at their phones anyway. All I know is nothing interesting will come of it. It’ll just melt into everything else. It feeds existing fires. You know, Donald Trump riding an eagle. It’s boring.

30:20 It’s the visual equivalent of auto-generated filler music for malls. That’s all it is. Forgettable stuff. You’ll hear it, move on, and want Panda Express. Collectively, the term “slop” locked in. It’s a perfect word. Forgettable strip mall garbage.

30:50 How do you get past slop? The good news is most people are not even there yet. We can relax for a minute. The fisheries industry not really all over AI. Your fish are free of slop right now. Not your code, though. You need systems and tools that enforce the constraints because then people will do good work. That is a common theme across everything.

31:30 Check us out at aboard.com. We make software that makes software using AI. We took a really different approach. We sort of taught it how to become a really good thoughtful partner and consultant. Try it at aboard.com. Send us an email at hello@aboard.com. We’d love to hear from you. We welcome all feedback. Unless it’s on iTunes where we really want you to give us five stars. Bye.