There’s a kind of framing that modern technology tells very well.
It tells you that the future is decided in boardrooms, venture rounds, and product launches. That the winners are determined by benchmarks, model performance, and enterprise contracts. That the most sophisticated product – cleanest, safest, most controlled-will inevitably win.
It’s a compelling story.
It’s also incomplete.
Because if you zoom out-beyond AI, beyond this moment, beyond the current hype cycle-you start to see a pattern. Not a theory. A pattern that has repeated across multiple generations of technology.
And that pattern says something much simpler:
The technologies that win are not the ones that are most refined.
They are the ones that become unavoidable.
History Doesn’t Reward Perfection. It Rewards Penetration.
We’ve seen this before. Not once. Repeatedly.
Google didn’t win because it monetised first.
It won because it answered better.
Before ads became the machine, before AdWords became one of the most powerful economic engines in history, Google was just a better way to find information. Faster. Cleaner. More relevant.
People didn’t choose Google because it had a business model.
They chose it because it worked.
Ads came later, almost as a consequence. And when they did, they didn’t feel like the product. They felt like a tax on something already essential.
That distinction matters.
Facebook didn’t win because it was sophisticated.
It won because it spread.
Not in the controlled environments of enterprise. Not through structured adoption cycles. But through people. Through identity. Through connection.
In many emerging markets, Facebook didn’t just become popular.
It became:
The default representation of the internet itself.
And once that happens, the question is no longer:
“Should I use this platform?”
It becomes:
“How do I operate without it?”
WhatsApp didn’t win because it had features.
If anything, it resisted them.
It won because it removed friction.
No ads.
No noise.
No complexity.
Just messaging that worked, everywhere.
And that simplicity allowed it to do something more powerful than growth.
It became infrastructure.
The Pattern Is Clear-But Often Ignored
If you strip away the branding, the narratives, the hindsight storytelling, the pattern looks like this:
1. Enter early
2. Solve a real problem
3. Scale rapidly (often free or subsidised)
4. Become embedded in behaviour
5. Monetise once switching becomes painful
Not theoretical.
Proven.
AI Is Following the Same Path. But With One Critical Difference.
AI is not just another application layer.
It is becoming:
• The interface to knowledge
• The interface to work
• The interface to thinking
And that changes the stakes.
Because this is no longer about:
Which product is better
It’s about:
Which system becomes default
The Conversation Today Is Too Narrow
Right now, the discourse around AI is still stuck in early-stage thinking.
• Model benchmarks
• Safety frameworks
• Enterprise deals
• Feature comparisons
These are not irrelevant.
But they are not decisive.
Because underneath all of this, a more important dynamic is unfolding:
AI adoption is being driven by curiosity, not procurement.
And curiosity scales differently.
Emerging Markets: The Deciding Layer Nobody Wants to Admit
There’s an implicit bias in how technology is analysed.
It assumes that:
• Developed markets lead
• Emerging markets follow
But in many cases, especially in the last decade, the reverse is true.
Emerging markets don’t just adopt technology.
They adapt it.
They stress-test it.
They scale it in ways that expose its real utility.
AI as a Leapfrog Mechanism
In developed markets, AI often feels like optimisation.
• Faster writing
• Better summaries
• Improved workflows
In emerging markets, it feels like access.
• Access to education
• Access to knowledge
• Access to capability
That distinction is not subtle.
It is structural.
Infrastructure Gaps Create Opportunity
Where systems are fragmented or underdeveloped, AI doesn’t compete.
It replaces.
• No tutor? → AI becomes one
• No consultant? → AI fills the gap
• No structured tools? → AI abstracts them
It doesn’t need to be perfect.
It just needs to be:
Good enough, accessible, and available
Tolerance Is Not the Same Everywhere
There’s a narrative-mostly originating from developed markets-that users demand:
• Clean interfaces
• No ads
• Perfect outputs
• Strong guarantees
That’s not universally true.
In many parts of the world, the equation is simpler:
Does this work for me?
If the answer is yes:
• Ads are tolerated
• Imperfections are accepted
• Trade-offs are understood
This Is Where Ads Become Misunderstood
There is a discomfort around ads in AI.
Understandably so.
Ads introduce:
• Bias risk
• Incentive misalignment
• Potential degradation of trust
But the conversation often stops there.
It doesn’t account for what ads actually enable.
Ads Are Not Just Monetisation. They Are Access Infrastructure.
Without ads:
• Free users = cost centres
• Usage must be constrained
• Growth is limited by burn
With ads:
• Free users = revenue contributors
• Usage can expand
• Access can scale
This is not theoretical.
This is how:
• Google scaled
• Facebook scaled
• Much of the internet exists
The Uncomfortable Truth About Sustainability
There’s a version of the future that assumes:
• Clean subscription models
• Pure enterprise revenue
• Controlled environments
It’s appealing.
But historically, it is not what scales globally.
What scales globally is:
A system that can fund its own growth through its own usage
Ads do that.
OpenAI: The Breadth Strategy
OpenAI, whether explicitly or implicitly, is building toward a very specific outcome:
Becoming the most widely used interface to intelligence
The Value of Messy Scale
OpenAI’s user base is not uniform.
It includes:
• Developers
• Enterprises
• Students
• Creators
• Casual users
• Curious users
On paper, this looks inefficient.
In reality, it is powerful.
Because:
The more diverse the input, the more adaptable the system
Learning From Reality, Not Structure
A system trained primarily on:
• Enterprise workflows
• Structured inputs
…becomes very good at predictable problems.
A system exposed to:
• Real-world queries
• Cultural variation
• Ambiguity
…becomes resilient.
That resilience matters more over time.
Anthropic: The Precision Strategy
Anthropic represents a different approach.
• More controlled
• More focused
• More enterprise-aligned
Strengths
• Higher signal quality
• Stronger trust positioning
• Better performance in structured domains
Trade-Offs
• Less exposure to global usage patterns
• Slower adaptation to cultural nuance
• Limited scale-driven learning
This Is Not a Competition of Models. It’s a Competition of Learning Systems.
And the difference is fundamental.
OpenAI
• Learns from breadth
• Prioritises scale
• Accepts noise
Anthropic
• Learns from precision
• Prioritises control
• Minimises noise
Both approaches produce intelligence.
But they produce different kinds of intelligence.
General vs Specialized Is the Wrong Debate
There’s a common framing:
• General models are shallow
• Specialized models are deep
At small scale, that holds.
At large scale, it breaks.
General Systems Evolve Into Specialized Systems
A system that:
• Sees more use cases
• Encounters more problems
• Processes more variation
…naturally develops specialised capabilities.
Not through design.
Through exposure.
The User Journey Reinforces This
Users don’t start with:
“I need an enterprise-grade AI solution”
They start with:
“Let me try this”
Then:
1. Curiosity
2. Habit
3. Dependence
4. Integration
At each stage:
• Value increases
• Stickiness increases
• Willingness to pay increases
The Role of Ads in This Journey
Ads don’t sit outside this system.
They support it.
Ads Enable the Top of the Funnel
• Free access
• High usage
• Continuous experimentation
Without this layer, the funnel narrows.
Ads Sustain the Middle
• Casual users remain viable
• Growth continues
• Engagement is maintained
Subscriptions and APIs Anchor the Bottom
• Power users pay
• Enterprises integrate
• High-value use cases stabilise revenue
The Real Risk Is Not Ads. It’s Misalignment
Ads themselves are not the problem.
Misaligned incentives are.
If the system optimises for:
• Engagement over accuracy
• Frequency over depth
• Revenue over trust
Then degradation happens.
The Google Lesson
Google didn’t fail because it had ads.
It risks decline when:
• Ads interfere with results
• Trust erodes
• Utility declines
The Same Will Apply to AI
If AI maintains:
High-quality output
Ads become tolerable.
If it doesn’t:
Users will look elsewhere
Where Business Value Quietly Takes Over
There is another layer often overlooked.
AI’s ability to:
Digest unstructured information
This Is Not a Feature. It’s a Shift.
Traditional systems required:
• Structure
• Inputs
• Defined workflows
AI can operate on:
• Documents
• Emails
• Notes
• Conversations
Messy data.
This Changes Enterprise Value
Tools like Copilot for Business (based on OpenAI models) demonstrate this clearly.
They don’t just:
• Generate content
They:
• Understand context
• Extract meaning
• Reduce cognitive load
This Is Where AI Moves From Tool to Capability
And this is where:
• Enterprise adoption deepens
• Revenue stabilises
• Long-term value is anchored
The Long-Term Prediction
If you remove hype, positioning, and narrative…
What remains is this:
1. Breadth Scales Faster Than Precision
The system that:
• Reaches more users
• Captures more behaviour
• Learns from more variation
…improves faster.
Not always cleaner.
But faster.
2. Ads Are Not a Weakness. They Are a Mechanism
They:
• Fund access
• Enable scale
• Sustain growth
History supports this.
3. Enterprise Value Will Consolidate Around Capability, Not Tools
The ability to:
• Understand unstructured data
• Operate across contexts
• Integrate into workflows
…becomes the differentiator.
4. Default Position Wins
The system that becomes:
• The default assistant
• The default interface
• The default tool
…wins behaviour.
And behaviour drives markets.
My Bet
The model that captures:
• The widest range of human thinking
• The most diverse set of inputs
• The most real-world usage
…has the advantage.
Because it is trained not on:
Ideal conditions.
But on:
Reality
Closing
There will always be:
• Cleaner systems
• More controlled models
• More precise outputs
But long-term dominance rarely comes from control.
It comes from alignment with how the world actually works.
And the world is:
• Messy
• Non-linear
• Context-driven
The system that absorbs that and still makes sense of it is not just useful.
It becomes:
Default
And once something becomes default,
it stops competing.
It defines the game.