Founder Mode Episode 52 - Grab A Shovel

Welcome back to Founder Mode!

In this episode, it was just Jason and me. No guest. No script.

We talked about something we’re both dealing with right now. AI is getting more powerful, but it’s also getting weirder.

Sometimes it helps a lot. Sometimes it confidently does the wrong thing.

And the tricky part is that the mistakes feel worse than human mistakes.

Let’s get into it.

1. AI Mistakes Feel Worse Than Human Mistakes

When a person messes something up, you expect it.

When AI messes something up, it feels different.

I explained it like this:

“It’s like I sent my kid to the store to get milk and they brought back orange juice.”

If you’re there with them, you fix it quickly. No big deal.

But if they leave, go to the store, pay, come back, and then you open the fridge, and it’s wrong, it feels way worse.

That’s what happens with AI agents.

They don’t just answer wrong. They complete the task incorrectly.

And by the time you notice, time is already lost.

2. AI Can Learn the Wrong Things

This one surprised me.

I ran an experiment where I asked AI to list everything it believed was true about my business.

Then I asked it to compare that with only what I had explicitly told it.

The result was wild.

“70% was wrong… it had recorded something as true that was only true to its own belief.”

It wasn’t a random hallucination. It was something more subtle.

It had created its own version of truth based on past conversations.

That means if you’re building on top of that, you could be building on shaky ground.

So now I treat AI memory like onboarding a new employee.

You don’t just let it learn everything. You guide what it remembers.

3. AI Should Be Treated Like a Junior Employee

This is the mental model that works best.

Not a tool. Not magic. Not perfect.

A junior employee.

You give it a task. It does the work. Then you review it.

You correct it. You refine it. You repeat.

If you expect perfection, you’ll be frustrated.

If you expect iteration, you’ll move faster.

The mistake most people make is skipping the review step.

They assume AI understands context the way a human does. It doesn’t.

At least not yet.

4. Systems Beat Prompts

A lot of people are still thinking in single prompts.

That’s not how this works anymore.

The real leverage comes from systems.

Jason walked through how his team builds content now:

One agent plans the structure
One writes
One edits
One creates images
One checks quality
One publishes

That’s a system.

Not one prompt.

That shift matters. Because once you think in systems, you stop asking “what can AI do?” and start asking “how do I design the workflow?”

5. Communication Is the New Moat

We ran a hiring process recently.

Over 2,000 applicants.

And most couldn’t clearly explain what they built.

That’s a problem.

Because in an AI world, your ability to communicate is everything.

You have to explain:

What you’re building
Why it matters
How it works
What the output should be

That applies to both humans and AI.

If you can’t explain it clearly, neither your team nor your agents will get it right.

“The right mental model is really like think of this as a new junior employee.”

That one idea fixes a lot of frustration.

5 Key Takeaways

AI mistakes feel worse than human mistakes because the task is already done when you catch them.
AI can create its own “truth” if you don’t manage its memory carefully.
Treat AI like a junior employee. Assign, review, and iterate.
Systems of agents outperform single prompts.
Communication is the most important skill in an AI-first world.