Before building anything, Tom says you can find out in about a day whether people care about the problem, and he explains how to run that test with AI.
A Note from James:
Back in the ’90s, I built websites.
I actually built American Express’s first website.
And I remember walking into places like JPMorgan and saying, “Eventually, every one of your customers is going to interact with you through a website.”
They laughed at me.
They weren’t stupid.
They just didn’t want it to be true.
I think about that a lot right now.
When I was a programmer, a good day might have been 100 lines of working, debugged code.
With AI, I can do something closer to tens of thousands.
Tom Bilyeu has spent his career reinventing himself.
He helped build Quest Nutrition. Then he built Impact Theory, which came out of nowhere and became one of the biggest shows in its category.
Now he’s building a video game called Kyzen—a massive, procedurally generated world—and using AI throughout the process.
So I asked him the question I was honestly a little scared to hear the answer to.
You’re 45. You’ve got a mortgage. You’ve got kids. You’ve built a career.
And AI is coming for your job.
What happens?
Tom did not comfort me.
He said history suggests some people will lose their jobs during a technological transition and never really recover.
That happens.
But then he made a distinction I think is much more useful.
It isn’t necessarily about age.
It’s about what kind of person you are when the world changes.
Do you look for the opportunity?
Or do you spend all your time looking at what you’re about to lose?
And Tom also gave me one question I’m absolutely stealing.
When he uses AI to evaluate an idea, he doesn’t ask:
“How do I make this succeed?”
He asks:
“Why is this going to fail?”
That changes the entire conversation.
This is Tom Bilyeu.
What You’ll Learn:
Timestamped Chapters:
[02:48] A Note from James
James connects the current AI transition to building websites in the 1990s and introduces Tom’s rule for challenging an AI-assisted idea.
[04:28] What Tom Is Building Now
Tom updates James on Impact Theory and introduces Kyzen, the massive procedurally generated game his team has been developing.
[05:39] What Is a Moat When Anyone Can Build Software?
James asks what protects a game—or any software company—once technical difficulty stops being a meaningful barrier.
[06:14] Brand and Network Effects Replace Technical Advantage
Tom argues that differentiation will increasingly come from audience, ecosystem, distribution, and the relationships surrounding a product.
[08:17] Winning in an Avalanche of Content
Impact Theory’s decade of competing on YouTube becomes Tom’s model for surviving a future where AI creates an overwhelming supply of games and media.
[10:04] The AI Wake-Up Moment
Tom remembers watching a game-development process that once required several people and months of work collapse into a single afternoon.
[11:51] Does AI Actually Create Better Businesses?
Tom says AI helps entrepreneurs get to revenue faster and with fewer people—but it does not automatically solve product-market fit.
[13:31] What Tom Would Do at 21
He lays out a practical approach to using AI as a researcher, product-development partner, and accelerator.
[15:40] The People-Pleasing Problem
James warns that AI can enthusiastically help build a terrible idea if the user never challenges it.
[16:24] “Why Is This Going to Fail?”
Tom explains how he prompts AI for disconfirming evidence, skeptical arguments, previous failures, and first-principles analysis.
[18:18] Use AI to Understand the Person Across the Table
Tom shows how founders can research distributors, customers, decision-makers, and the incentives driving them.
[19:25] Better Inputs Create Better AI Outputs
A frustrating design exercise teaches Tom that vague stylistic references produce derivative results; understanding the underlying vocabulary produces more original work.
[23:12] When Everything Becomes AI-Native
Tom predicts that the next stage will come when software, appliances, development environments, and everyday tools are built around AI from the beginning.
[25:40] Your Old Advantages Just Disappeared
Tom describes the unsettling realization that intelligence, speed, technical skill, and capital may no longer provide the same protection they once did.
[27:24] From 100 Lines of Code to 50,000
James compares his early programming career with the amount of work modern AI tools can generate.
[28:25] Today’s AI vs. Artificial Superintelligence
James and Tom distinguish the limitations of current models from the far more disruptive future systems AI skeptics and safety researchers worry about.
[30:00] What Happens to the 45-Year-Old Worker?
James asks the question directly: if you’re mid-career with a family and AI threatens your job, what happens next?
[32:01] Money Doesn’t Solve the Meaning Problem
Tom argues that even if governments eventually cushion widespread unemployment, people will still need purpose and identity.
[33:10] Be the Last Human Standing
As long as AI remains a tool, Tom says the advantage goes to the human who learns how to use it aggressively.
[34:16] What Should You Learn Right Now?
Tom recommends experimenting with the strongest available AI tools toward work that is economically valuable.
[36:17] James Remembers the Early Internet
Companies once laughed when James told them customers would eventually interact through websites. He sees a similar resistance happening now.
[37:18] Reinvention Is a Personality Trait
Tom argues that disruption rewards people who become excited by learning the new thing instead of focusing exclusively on what they may lose.
[38:00] Is the AI Infrastructure Boom a Bubble?
James and Tom discuss falling inference costs, enormous data-center investment, debt, and the historical pattern of infrastructure booms.
[42:35] Even If the Bubble Bursts, AI Still Wins
Tom argues that AI is already useful enough to spend years transforming existing industries even if model progress stopped tomorrow.
[43:11] Which Jobs Last the Longest?
Physical work may resist automation longer, but Tom sees embodied AI and robotics eventually reaching even highly skilled trades.
[45:48] AI, Medicine, and the Human Layer
James uses a medical story to explore the difference between AI’s analytical ability and the continued need for human judgment, care, and communication.
[47:53] How Impact Theory Broke Into a Crowded Market
James asks how Tom built a major media brand after being told podcasting was already saturated.
[48:37] There Is Always Room for Better
Tom explains the principle behind Impact Theory’s early growth: find something valuable competitors are not doing and execute it at a higher level.
[50:24] Find the White Space
High production value helped Impact Theory attract guests before it had a large audience. Tom says every new entrant needs to identify its own equivalent advantage.
[51:31] Building Media With AI
Tom describes using AI for analytics, trend detection, scripting, content generation, and marketing strategy around Kyzen.
[53:39] Testing Kyzen Before the Public Sees It
Tom explains why the team is starting with trusted testers before expanding access—and why brutally honest feedback is welcome.
[55:35] Why It’s Called Kyzen
The name comes from the Japanese concept of continuous improvement, deliberately respelled so the project could be trademarked.
[56:18] Constant Improvement
James connects Kyzen’s name back to Tom’s career of repeated reinvention and closes the conversation.
Additional Resources: