Every major technology shift produces two things at the same time: real value and exaggerated expectations.
AI is no different.
The question founders should be asking isn’t “Is AI a bubble?” It’s “Where is the bubble, and where is the durable value?”
Recent reporting by Derek Thompson (The Atlantic) offers a useful lens for separating the two.
What the AI Bubble Actually Looks Like
Insights inspired by Derek Thompson’s analysis
Thompson’s core observation is not that AI is fake, but that the narrative around AI has outrun its current economic reality.
A few key points he highlights:
AI adoption is widespread, but revenue capture is narrow
Productivity gains are real in specific tasks, not across entire organizations
Many companies are “AI-washing” existing products to ride attention cycles
Capital is flowing faster than proven business models
In other words, usage does not equal sustainable value.
This is a classic early-cycle pattern.
Where the Hype Concentrates
The bubble risk is highest in areas where:
AI is positioned as a replacement for judgment rather than labor
Products promise transformation without workflow change
Valuations are justified by demos instead of distribution
As Thompson notes, transformative technologies often take longer to reshape productivity than headlines suggest, even when they eventually succeed.
Steam engines, electricity, and the internet all followed this arc.
AI is likely doing the same.
Where the Real Value Is Quietly Compounding
The most durable AI businesses today share a few characteristics:
They sit inside existing workflows
They replace specific tasks, not entire roles
They are narrow, operational, and unglamorous
They charge for outcomes, not intelligence
This aligns with Thompson’s broader argument: general-purpose technologies don’t create instant revolutions. They create slow, uneven gains that reward disciplined builders.
What This Means for Founders
If you’re building with AI, the takeaway isn’t to retreat. It’s to be precise.
Strong founder positioning today looks like:
Fewer promises and tighter scope
Less “AI-powered” and more “time saved” or “errors reduced”
Measurable ROI instead of speculative narratives
Founders who survive hype cycles don’t fight them. They build underneath them.
A Founder’s Mental Model
Think of AI as:
A new layer of infrastructure
Cheap cognitive labor
Unreliable by default, powerful when constrained
The opportunity isn’t to predict the future. It’s to build systems that work even if the hype resets.
Closing Thought
AI may be in a narrative bubble. That doesn’t mean it’s a business bubble.
As Derek Thompson’s reporting reminds us, technologies don’t fail when expectations are high. They fail when builders confuse attention for value.
The founders who win this cycle will be the ones who stayed quiet, specific, and useful.
Inspired by reporting and analysis from Derek Thompson, The Atlantic Written for founders building through noise, not chasing it.