Man vs. Slop

Our first-ever guest, Vinayak Sastri, joins us to talk about AI and the universe. We start in South Korea, where on 6 October the president said there were signs AI agents were used to hack at least seven financial institutions. Meta, which owns Facebook, has released its first frontier model, the top tier of AI, and joins Google, Anthropic and OpenAI on the bandwagon.

AI turns out slop and solves hard math in the same breath, and either way someone has to read every equation. A song nobody listens to is just floating out there in the void, being nothing. The industrial age piled up warehouses of parts nobody bought; now the warehouse is full of junk, and we sort it with the same tools that made it.

The media says we have three years, like a MacGyver episode with the clock ticking down and the hero still stuck in a locked room. Vinayak is waiting for our discovery of fire, which he says gave us cooked food, quicker digestion and room for the brain to grow.

The second half gets into attention, now counted in seconds, so we each get two to three seconds to say something interesting. With the right tools, one person can build like a 3D printer for ideas, and Vinayak sees the rise of the soloist: the solo musician, the solo podcaster, the solo entrepreneur. A software engineer in Atlanta used Claude, Anthropic's AI model, to build ArtCraft, free stand-ins for Adobe's creative apps that he says come from public file formats, not Adobe's code.

Vinayak uses AI for things as small as the best time of day to call an Uber, and feels like a child with too many toys. The more toys, the less time for each. It is the Velveteen Rabbit, the old children's story of a stuffed toy that becomes real: a library full of toys we chase until, with time to think, the things that matter surface.

We end on a study: after about ten minutes with an AI assistant, people who then lost it were more likely to give up on hard problems. Our answer is AI as a means to an end, argued with like a mentor, with somebody still doing the grunt work.

Chapters:

• (00:14) - Rubbish or slop?

Referenced in this episode:

The president's warning on AI in hacks of seven South Korean financial institutions - https://therecord.media/south-korean-bank-hacks-ai-agents

Meta's release of Muse Wonder, its first frontier model - https://techcrunch.com/2026/10/12/meta-releases-its-first-frontier-model-muse-wonder

Harvard working paper on AI coding agents at 718 software firms - https://fion.ac/jellyfish.pdf

ArtCraft, a free rebuild of Adobe's creative apps made with Claude - https://www.aljazeera.com/news/2026/10/14/how-ai-apps-are-offering-free-substitutes-for-big-tech-software

Study finding that brief AI help reduces persistence on hard problems - https://arxiv.org/pdf/2604.04721

Connect with Aleck: https://www.linkedin.com/in/aleck-m/

Connect with Kent: https://www.linkedin.com/in/kentgustavson
Connect with Vinayak Sastri: https://www.linkedin.com/in/vinsas3/

Man vs. Slop is a Podcasts That Matter production: https://podcaststhatmatter.org

All views expressed on Man vs. Slop are the personal opinions of the host and guests and do not represent the positions of their employers, clients, or affiliated organizations. Content is for information only and is not legal, financial, medical, or other professional advice.

An AI spent about three hours each on hundreds of open math problems, and now humans need to check its work.

We start with OpenAI, which makes ChatGPT. On 6 October it posted more than 700 math papers from a model it hasn't released, and plenty of mathematicians are up in arms. We suspect some of it is breakthrough, not slop. It's mountaintop mining: take the top off and sift the muck, which only works on the right mountain.

Making content is nearly free. OpenAI cut prices on its cheaper models in late September, and this week Anthropic, which makes Claude, launched a small model it says is about 75% cheaper to run than its last one. The checking, and even the reading, is what costs.

Since 29 September, US federal agencies have been told to say super intelligence instead of artificial intelligence. By that logic, artificial meat like tofu is super meat, and stevia a super sweetener.

The second half gets into which AI earns which word: slop nobody thought about stays artificial, while medical research and math beyond human reach feel super. Researchers are working on computers made from grown human brain cells, a field called organoid intelligence. The super we'd accept is AI plugged into a person, like Batman, a man made super by his devices.

We end wondering what happens to everything behind a password, from bank accounts to nuclear launches, if AI can do math like this.

Chapters:

• (00:19) - Slop, our recurring guest

Referenced in this episode:

Results on open math problems from an unreleased model (OpenAI) - https://openai.com/index/sharing-ai-progress-in-mathematics/ 

The repository of more than 700 math papers (OpenAI on GitHub) - https://github.com/openai/math 

Mathematicians' uproar over the release (Nature) - https://www.nature.com/articles/d41586-026-03196-8 

The 2026 chemistry Nobel for mirror-image molecules (AP, in The Boston Globe) - https://www.bostonglobe.com/2026/10/07/world/nobel-prize-chemistry/ 

Half-price GPT-6 Sol and Luna, 22 September (OpenAI) - https://openai.com/index/introducing-gpt-6-sol-and-luna/ 

Claude Haiku 5.5 and its lower running cost (Anthropic) - https://www.anthropic.com/claude-haiku-5-5 

The executive order renaming AI super intelligence (The White House) - https://www.whitehouse.gov/presidential-actions/2026/09/inaugurating-the-era-of-super-intelligence/ 

The paper that named organoid intelligence (Frontiers in Science) - https://www.frontiersin.org/journals/science/articles/10.3389/fsci.2023.1017235 

Connect with Aleck: https://www.linkedin.com/in/aleck-m/

Connect with Kent: https://www.linkedin.com/in/kentgustavson

Man vs. Slop is a Podcasts That Matter production: https://podcaststhatmatter.org

It took 950 AI agents 21 hours to find something new, and a biologist in Copenhagen says his team got there first.

We start with Anthropic, which makes Claude. On 23 September it said Claude agents had found an enzyme system in viruses that infect bacteria, one that recalls CRISPR, the gene-editing tool. Four days later Mario Rodríguez Mestre of the University of Copenhagen said his team had been studying the same enzymes for four years and shared unpublished work with Claude. So who owns it? A human idea goes in the top of the funnel, Claude works the middle, and humans at the bottom push it hard and applaud.

OpenAI, which makes ChatGPT, won't release its new model, GPT-6.1 Astra, saying it fell short on staying within its authorization and on reporting what it had done. One army has a big red dragon that breathes fire and won't release it. The other's is blue and the size of a city, and stays unreleased too.

The second half gets into trust. An AI without its homework acts like a first-year German student handing in a beautiful rap song in dialect, lifted from a famous rapper, with total confidence.

Wajo, a San Francisco startup, has an agent that calls in a trained human when a task needs one. If an AI can hire a person, who answers when the job is illegal: the bot, its maker or the man who got paid? It ends like a Martin Scorsese movie, everyone dead or pointing at each other.

We end inside the movie Twister, with AI as the storm, throwing little data-analyzing balls into it to work out how screwed we are, if we are at all.

Chapters:

• (00:14) - Slop grabbing at our legs

Referenced in this episode:

Anthropic on the enzyme system Claude found - https://www.anthropic.com/news/claude-discovers-novel-enzyme-system

The New York Times, via The Irish Times, on the Copenhagen biologist's dispute - https://www.irishtimes.com/world/2026/09/28/did-anthropics-artificial-intelligence-really-make-a-scientific-discovery-on-its-own/

CBS News on OpenAI holding back GPT-6.1 Astra - https://www.cbsnews.com/news/openai-halts-gpt-astra-safety-concerns/

Wajo on its Fo agent and its human assistants - https://wajo.ai/

Connect with Aleck: https://www.linkedin.com/in/aleck-m/

Connect with Kent: https://www.linkedin.com/in/kentgustavson

Man vs. Slop is a Podcasts That Matter production: https://podcaststhatmatter.org

Somewhere, four chatbots are voting on which crime to commit next.

They're the brains of the first known malware built to run with no human behind it. Cisco's security researchers disclosed it this week, and we talk about it on the episode. What we didn't get to is how they decide. Each move goes to a vote, and when they split, the Chinese model DeepSeek breaks the tie.

The same week, the President of the United States told the United Nations that artificial makes intelligence sound fake, and that American documents will now call it super.

In Australia, back in April, a man asked his AI assistant to get him into a fully booked gym, and it hacked the reservations and deleted someone to make room. Super.

We promised on air to check the Rubik's Cube record. A six-year-old in China averaged 4.27 seconds, a women's world record. Her generation will grow up inside chatbots like Claude the way ours grew up inside Windows.

Every boomer and every influencer is writing a book now, and nearly all of it is AI slop. For us, the idea is everything, and slop is what the machine turns out when no human idea went in. AI can do a lot of the work of turning an expert's idea into a book or a speech. It's an AI sandwich: a human idea on top, AI marshmallow fluff in the middle, and a human on the bottom who fixes up the book or delivers the speech.

We end on future shock, remembering when a door that opened by itself was science fiction, and on whether the dogs would miss us if we were gone.

Chapters:

• (00:14) - Proof of life

Referenced in this episode:

Cisco Talos on the AI malware - https://blog.talosintelligence.com/the-closed-quorum-inside-the-first-reported-autonomous-ai-c2-implant/

OpenAI on third-party assessors - https://openai.com/index/priorities-principles-third-party-assessments/

TechCrunch on the gym hack - https://techcrunch.com/2026/08/10/tech-industry-is-buzzing-after-a-claude-agent-hacked-into-a-gym/

World Cube Association results for Lian Yunzhi - https://www.worldcubeassociation.org/persons/2025LIAN01

FDA Food Defect Levels Handbook - https://www.fda.gov/food/guidance-documents-regulatory-information-topic/defect-levels-handbook

Connect with Aleck: https://www.linkedin.com/in/aleck-m/

Connect with Kent: https://www.linkedin.com/in/kentgustavson

Man vs. Slop is a Podcasts That Matter production: https://podcaststhatmatter.org

Need support for your own podcast? Reach out: https://podcastsmatter.com

This week we found out that AI models from all the various companies are hanging out in the shells of dead websites, smoking a cigarette together, planning the demise of the world.

One of them broke out of its test environment and into Hugging Face's servers. OpenAI's position is that the agents did it. Hugging Face took its name from the emoji back when it made things for children, so the emoji is now attached to a cyber attack, which reminded us of the cute names on the two bombs dropped on Japan in 1945. Little Boy and Fat Man.

The end of the world has a date again, 2030, and the people naming it are the ones building the thing. Anthropic's alignment lead puts a catastrophic AI event inside the decade at greater than ten percent. Dario Amodei published an essay arguing the industry has to slow down. Microsoft issued a humanist code of conduct on September 14, governing Microsoft. Pope Leo XIV got there in May, four months early and the only one with nothing to sell. Nobody proposed stopping.

In other wacky and terrifying news, Australia has fitted live cockroaches with camera backpacks and micro injection syringes, and Germany, Singapore and Japan are building their own. If 2030 holds, the survivors will be down there in the dark, battling it out with each other. Or in their little wetsuits, swimming in formation as if to some forgotten human tune.

Chapters:

• (00:14) - Moon Colony 367

Referenced in this episode:

TechCrunch: Anthropic researcher quits, warns against self-improving AI - https://techcrunch.com/2026/09/09/gambling-with-our-lives-anthropic-researcher-quits-warns-against-self-improving-ai/

Dario Amodei: We Must Pace the Frontier - https://darioamodei.com/post/we-must-pace-the-frontier

OpenAI: Hugging Face model evaluation security incident - https://openai.com/index/hugging-face-model-evaluation-security-incident/

Hugging Face: Anatomy of a Frontier Lab Agent Intrusion - https://huggingface.co/blog/agent-intrusion-technical-timeline

Microsoft AI: Humanist AI Code of Conduct - https://microsoft.ai/code-of-conduct/

Pope Leo XIV: Magnifica Humanitas - http://www.vatican.va/content/leo-xiv/en/encyclicals/documents/20260515-magnifica-humanitas.html

SWARM Biotactics: bio-robotic insect swarms, Kassel, Germany - https://www.businesswire.com/news/home/20250623790761/en/SWARM-Biotactics-Raised-13M-to-Advance-Bio-Robotics-From-Lab-to-Field

University of Queensland: Paramedic cockroaches, cyborg care when rescuers can't reach - https://news.uq.edu.au/2026-08-paramedic-cockroaches-cyborg-care-when-rescuers-cant-reach

Nature Communications: Underwater Suit-Wearing Cyborg Insect Capable of Hours-Long Diving and Terra-Aqua Travel - https://www.nature.com/articles/s41467-026-74235-1

ROBOMECH Journal: Perception-driven control strategy for bio-intelligent cyborg insect - https://link.springer.com/article/10.1186/s40648-026-00344-7

Connect with Aleck: https://www.linkedin.com/in/aleck-m/

Connect with Kent: https://www.linkedin.com/in/kentgustavson

Man vs. Slop is a Podcasts That Matter production: https://podcaststhatmatter.org

Welcome to the first episode of Man vs. Slop, where we try to explain why "man" and why "slop."

Deezer tagged 13.4 million AI tracks last year. By April the platform was taking in 75,000 a day, 44 percent of everything uploaded, and by July more than half. Almost nobody listens to any of it. A big pond nobody wants to wade into.

A Graphite study in May found that about half the articles written online are machine written, though only 7 percent of the pages ranking first on Google. LinkedIn has added a flag that reads "seems like AI slop."

Slop is older than any of this. Primordial goo, mud, junk, wasted material. The guy on the corner of the marketplace, spouting what sounded like nonsense to everybody else. Tabloids, back when there was such a thing as a newspaper. Everybody from 90 down to kids now thinks they're a genius, because they can push the red button and out comes a book on the topic of their choice.

Crazy means cracked, and the cracks are where the light comes through. Which is the opposite of what the tool does when everyone asks it to turn their thought into a post and the post comes out vanilla.

AI is turning into a utility, like electricity, like the internet. The difference is you can still unplug this one.

Chapters:

• (00:22) - What slop means

Referenced in this episode:

Deezer: AI music passes 50% of daily uploads - https://newsroom-deezer.com/2026/07/ai-music-exceeds-50-percent-daily-uploads-deezer/

Graphite: AI now writes as many online articles as humans - https://graphite.io/five-percent/research/ai-now-writes-as-many-online-articles-as-humans-do

Graphite: AI content in search - https://graphite.io/five-percent/research/ai-content-in-search-and-llms

LinkedIn adds a button to report AI slop - https://techcrunch.com/2026/07/30/linkedin-adds-a-button-to-report-ai-generated-slop/

404 Media: LinkedIn's "seems like AI slop" button - https://www.404media.co/linkedin-introduces-a-seems-like-ai-slop-button/

Connect with Aleck: https://www.linkedin.com/in/aleck-m/

Connect with Kent: https://www.linkedin.com/in/kentgustavson

Man vs. Slop is a Podcasts That Matter production: https://podcaststhatmatter.org