What Naval Missed
In 2018, Code and Media were leverage. In 2026, they're commodities. The moat moved.
What Naval Missed
Naval Ravikant’s thesis was elegant: find your specific knowledge, attach leverage (Code or Media), escape competition through authenticity. In 2018, this was revolutionary. Build software. Create content. Scale without permission.
In 2026, an LLM writes the software and generates the content. The leverage Naval described is no longer scarce. The question is no longer “can you produce Code and Media?” The question is “can you govern the agents that produce them for you?”
Familiar Ground
The leverage menu has not changed. Labour, Capital, Code, Media. Naval was right that Code and Media are the permissionless forms: no gatekeepers, no investors needed, infinite scale at zero marginal cost.
What changed is who produces them. In 2018, writing software required years of training. Creating content required editorial skill. Both were hard enough to be differentiators. The founder who could code AND write had a compounding advantage.
In 2026, a solo founder with the right prompts produces more code in a week than a ten-person engineering team produced in a month three years ago. The production barrier collapsed. Code and Media are no longer leverage. They are table stakes.
Counter-Signal
VCs have not caught up. They are still funding headcount lines on budgets that AI has already replaced. “We need $3M to hire 15 engineers.” In 2026, those fifteen engineers are prompts. The cost is not a series of salaries. It is a series of API calls.
This creates a multi-year window. The venture capital paradigm assumes that scale requires people. AI proved that scale requires orchestration. The solo founder who understands this builds a moat while the funded competitor builds a burn rate.
But here is what Naval’s framework misses: orchestration without governance is chaos. A solo founder running twenty AI agents without a decision framework, without quality gates, without audit trails, is not leveraged. They are drowning in unverified output.
⚛️ The Fusion
Two eras crash here, and the collision reveals a missing layer.
The 2018 thesis says leverage is Code and Media. Build software that works while you sleep. Create content that compounds without your presence. This was true and remains structurally true: zero marginal cost is still zero marginal cost.
But the thesis assumed that producing Code and Media was the hard part. It was. Now it is not. An LLM produces both. The hard part moved.
The 2026 update says the hard part is governance. When your agents can produce unlimited Code and Media, the bottleneck is not production. It is curation: which output ships, which gets rejected, how conflicts between agents resolve, what quality means, and who (or what) enforces it.
The missing layer in Naval’s framework is the Operating System:
| 2018 (Naval’s Framework) | 2026 (AI Update) |
|---|---|
| Specific knowledge is the moat | Specific knowledge is still the moat, but now it governs agents, not skills |
| Code is leverage | Code is commodity (AI writes it) |
| Media is leverage | Media is commodity (AI generates it) |
| The founder produces | The founder curates and governs |
| Scale = zero marginal cost production | Scale = zero marginal cost production + governance |
| Missing: nothing | Missing: the OS that governs the agents |
The governance thesis is the collision. The first solo unicorn will not be the best programmer. It will be the best systems orchestrator. The founder who builds the governance harness (decision frameworks, quality gates, multi-agent dialectics, audit trails) will outperform the founder who writes more code.
Why? Because ungoverned agents produce at scale but compound errors at scale too. A thousand lines of unreviewed code is not an asset. It is technical debt at machine speed. A hundred articles with no editorial standard is not content marketing. It is noise.
Note: Naval’s leverage framework updated for 2026: Code and Media, once scarce leverage, are now commodities. The new moat is the Operating System that governs the agents producing them.
The New Pattern
The practical question for every solo founder: do you have an OS, or are you just running agents?
The OS includes three things Naval’s framework did not need in 2018:
Decision governance: when two agents disagree on the right approach, how does the conflict resolve? Without a protocol, the founder becomes the bottleneck, manually reviewing every output. With a protocol (structured dialectic, adversarial review, adjudication criteria), the system governs itself.
Quality gates: when an agent produces output, how do you know it meets the standard? Without gates, the founder ships whatever the agent produces. With gates (automated audits, multi-agent review, standard compliance checks), quality compounds instead of degrading.
Knowledge accumulation: when an agent learns something in one task, how does that knowledge persist to the next task? Without a knowledge system, every agent session starts cold. With one, the OS gets smarter over time.
Naval was right about the destination: productise yourself. The route has changed. In 2018, you productised by building Code and Media. In 2026, you productise by building the OS that governs the agents that build Code and Media.
The founder who builds the harness owns the compound. The founder who skips the harness produces at speed and drowns in the output.
The Open Question
Naval said Code and Media are the permissionless leverage. He was right in 2018.
In 2026, everyone has Code and Media leverage. The AI gave it to them for free.
What leverage do you have that the AI cannot replicate?
This fusion emerged from a STEAL on the 2026 update to Naval’s “Productize Yourself” framework, tracing the shift from production leverage to governance leverage.