S2E32: Yelling at Robots: The Mindset Shift Every Business Leader Needs to Have About AI

The AI conversation is everywhere. LinkedIn, Substack, your team meetings, your inbox. But most of what you’re hearing is either hype or doom. Tech executive, AI builder, and “Yelling at Robots” Substack writer Chris Perrin joins Stacie today to cut through both.

What you’ll hear in this episode:

  • The cult of personality behind OpenAI, Anthropic, and Grok, and why it should matter to you

  • What AI agents actually are (hint: it’s not magic, it’s math)

  • The swarm disaster that took 6 days and cost Chris everything he built, only to redo it in 4 hours

  • Why vibe coding is exciting AND why you still need a human who can call BS

  • The “AI slop” problem and what happens when leaders hand responsibility to the robot

  • What superintelligence actually requires, and why we’re not there yet

  • How humans and AI can actually collaborate, and what that looks like in practice

  • The one thing that separates a good AI output from a great one (spoiler: it’s you)

The People Problem

Chris’s Substack isn’t called “Yelling at Robots” because he hates AI. He loves it. He uses it every day. He even asks Claude to fact-check his complaints before he publishes them.

The name exists because the real problems in AI aren’t technical. They’re human.

Sam Altman. Dario Amodei. Elon Musk. Each one of these founders has built something extraordinary. Each one is also playing politics, lobbying governments, and shaping public narrative in ways that benefit their companies first and the rest of us second. Chris calls it what it is: a cult of personality. And he thinks we should all be paying closer attention to who’s behind the model we’re using, and why.

What AI Actually Is (And Isn’t)

Here’s the part most people skip. AI is math. Very sophisticated math, but math.

When you type a prompt, the model takes your words, converts them to numbers, runs calculations to predict the most likely next token (word), and converts that back to language. That’s it. There’s no thinking. No feeling. No understanding of the mystery of life.

What makes models different from each other comes down to weights: the probabilities assigned to what word should come next. And those weights are shaped by the people who trained them. Which is exactly why Chris thinks the “who’s behind this model” question matters more than most people realize.

The Swarm Disaster (And What It Taught Him)

Chris wanted a swarm. He’d seen other AI builders talk about them, mapped out the workflow, built the agents. He had an architect agent, coder agents, a QA agent. He had a system.

What he ended up with was a QA agent that spawned 13,000 unit tests and a process that ran for 6 days straight, going nowhere.

The problem? The agents were written in a way that meant they would never be satisfied. Stuck in a loop, burning tokens, referencing a shared context file that kept getting misread. When he deleted everything and started over, the same task took 4 hours.

Swarms are easy to build wrong and very hard to debug when you do. The complexity hides the failure.

AI Slop Is Real, And It’s Your Fault

AI will tell you the output is good enough. It won’t tell you whether it actually is.

That space between “the robot said it’s fine” and “it’s actually fine” is where human expertise lives. Chris calls the result of skipping it “AI slop.” It’s what happens when leaders hand over responsibility to the model and walk away. The output looks plausible. It might even look polished. But it isn’t staked on anyone’s reputation, and it shows.

The fix isn’t to use less AI. It’s to build teams that can call BS on the output. People who know enough about the problem to examine the answer. That’s the collaboration model that actually works.

Can AI Replace Your Whole Team?

Technically? Probably yes, eventually.

Practically? Not without someone who understands the problem well enough to know if the answer is right.

Chris draws a line between three things we used to call “programming” but are actually separate: generating raw code, solving specific problems, and system-level thinking. That third one, knowing what problem to solve and why, is still a human job. And it’s true in every field: law, marketing, medicine, copywriting. The doing can be automated. The deciding still can’t.

Why This Matters For You

If you’re a CMO, a marketing leader, or a founder using AI tools in your business, this episode is the grounding you need.

You don’t have to be a coder to use AI well. But you do have to understand the problem well enough to evaluate the output. You have to be the person who can say “wait, that’s not right.” You have to stay in the loop.

Because the leaders who hand it all to the robot and walk away? They’re building on a foundation of AI slop. And that catches up with you.

Where to Find Chris

stormsoftwaregroup.com | yellingatrobots.substack.com | LinkedIn | X/Twitter | Medium

Next
Next

S2E31: The Orange Flag: What AI Agents Going Rogue Tells Us About the Superintelligence Race