From “Read Write Own” to AI: Who Will Control the Intelligence Layer?

The internet has gone through distinct phases. First, we read. Then, we wrote. Now, according to Chris Dixon’s thesis, we are entering an era where users can own.

But what happens when this framework is applied to artificial intelligence?

AI is not quite the same as internet, it is rapidly becoming the interface to it. That raises a question:
Will AI follow the same trajectory toward user ownership, or will it consolidate power even further?

The Parallel: Mapping “Read Write Own” to AI

1. The “Read” Phase → AI as Consumption

Today’s AI experience is largely passive:

• You ask questions

• The model responds

• You consume the output

Users don’t own:

• The models

• The training data

• The infrastructure

This mirrors the early web, where users were primarily consumers of static content.

2. The “Write” Phase → AI as Creation

We are quickly shifting into a more participatory phase:

• People build prompts, agents, and workflows

• Developers fine-tune models and create AI-powered apps

• Entire businesses are emerging around AI-generated content

But there’s a catch.

Just like Web2:

• Platforms control distribution

• APIs gate access

• Value accumulates centrally

You can create with AI but you don’t fully own what you create.

3. The “Own” Phase → Still Unwritten

This is where things get interesting and uncertain.
What would it mean to “own” AI?

Possibilities include:

• Owning your personal AI agent

• Controlling your data and how it’s used for training

• Sharing in the economic upside of models you contribute to

• Moving your AI identity across platforms

But unlike crypto, AI doesn’t map cleanly to ownership:

• Models are probabilistic, not discrete assets

• Training data is deeply entangled

• Attribution is nearly impossible at scale

• Compute requirements favor centralization

In other words: AI resists clean ownership models.

The Real Shift: From Ownership to Control

Instead of “Read → Write → Own,” AI may follow a different arc:

Access → Create → Control

Control may be more realistic than ownership.

This includes:

• Control over your data

• Control over your AI assistant

• Control over how your outputs are used

Rather than owning the model itself, users may own their interface to intelligence.

Where This Is Already Heading

Several emerging patterns hint at what the “control layer” might look like:

Personal AI Agents

AI systems that:

• Learn from your behavior

• Persist over time

• Act on your behalf

These could become your primary digital asset: not something you rent, but something that represents you.

Local and Edge Models

Running models locally:

• Reduces dependency on centralized providers

• Gives users more privacy and autonomy

Though still limited, this trend is accelerating.

Data Ownership Conversations

There is growing pressure around:

• Who owns training data

• Whether contributors should be compensated

• How provenance is tracked

This remains an unsolved problem but a critical one.

The Central Tension

AI has strong gravitational pull toward centralization:

• Massive compute requirements

• Data network effects

• Talent concentration

At the same time, there is increasing demand for:

• Transparency

• Portability

• User control

The future of AI will likely be shaped by this tension; not resolved by it.

So, Will AI Follow “Read Write Own”?

Partially.

• We are clearly moving from Read → Write

• The Own phase is not guaranteed

• What emerges instead may be a hybrid:

  • centralized infrastructure

  • user-controlled interfaces

The most important shift may not be owning AI systems but not being owned by them.

Closing Thought

If the last era of the internet was about owning content,
the next era may be about owning or at least controlling intelligence itself.

And that distinction will define who captures value in the AI age.