The Amplification of Bottlenecks

The smartest model in the world is useless without the right context.

— Mike Krieger

When you make one part of a system dramatically faster, you don’t just speed up the whole system.

You reveal where it was actually broken.

At Anthropic, 90% of code is now written by AI. Engineering—once the primary constraint—has been obliterated as a bottleneck.

What emerged in its place?

The New Constraints

Decision-making. What should we build? Who decides? How do we align? When code generation becomes instantaneous, the time spent debating requirements and priorities suddenly dominates the development cycle.

Integration. The merge queue collapsed under the weight of AI-generated pull requests. Code review processes, designed for human-paced development, crumbled under the volume. Traditional CI/CD pipelines became the chokepoint.

Context. The difference between knowing your internal documents, Slack conversations, and domain expertise versus starting from scratch is “entirely the difference between a good answer and a bad answer.”

This is the pattern of progress: solve one constraint, amplify the next.

Historical Echoes

The printing press didn’t just make books faster to produce. It revealed that literacy was the real bottleneck. Suddenly, the ability to read became more valuable than the ability to physically copy text.

The internet didn’t just make information faster to access. It revealed that attention was the real bottleneck. The limiting factor shifted from information scarcity to information filtering.

Steam engines didn’t just make transportation faster. They revealed that logistics and supply chains were the real constraints on industrial growth.

Each breakthrough follows the same pattern: eliminate the obvious limitation, discover the hidden one.

The Context Revolution

AI doesn’t just make coding faster. It reveals that clarity is everything.

Mike Krieger’s equation captures this perfectly:

Model Intelligence + Context & Memory + Interface = Utility

From his recent guest appearance on Lenny’s Podcast.

The competitive advantage doesn’t come from having the best AI—it comes from giving AI the best context.

What This Means for Organizations

Every company using AI will face this amplification effect. The question isn’t whether it will happen, but which constraint will emerge first.

For Software Companies:

• Coding speed → Decision paralysis
• Feature development → Product strategy alignment
• Technical implementation → User research and validation

For Content Companies:

• Writing speed → Editorial judgment
• Content generation → Audience understanding
• Production volume → Distribution effectiveness

For Research Organizations:

• Data analysis → Question formulation
• Literature review → Hypothesis generation
• Methodology execution → Interpretation skills

The Preparation Problem

Most organizations aren’t ready for their new bottlenecks.

They’re still optimizing for the old constraint. Hiring more engineers when the real need is better product managers. Investing in faster hardware when the real need is clearer communication protocols.

The winners will be those who anticipate the amplification.

Instead of just implementing AI tools, they’ll ask: “When this constraint disappears, what becomes the new limitation? How do we strengthen that now?”

The Meta-Pattern

This reveals something deeper about progress itself.

Every breakthrough isn’t just an acceleration—it’s a reorganization of constraints. The system doesn’t just get faster; it gets fundamentally different.

The organizations that thrive aren’t those that get the best AI tools first. They’re those that redesign their systems around the new constraint landscape.

When everyone can generate code instantly, competitive advantage flows to those who know what code to generate and why.

When everyone can create content at scale, advantage flows to those who understand what content matters and for whom.

When everyone can analyze data automatically, advantage flows to those who know which questions to ask.

The Strategic Question

The question isn’t whether AI will make your work faster.

The question is: what bottleneck will it reveal in your organization?

And are you ready for it?

The organizations that answer this question correctly—and prepare accordingly—will have overwhelming advantages in the AI-native economy.

The rest will find themselves optimizing for constraints that no longer exist while struggling with limitations they never saw coming.

---

Every breakthrough amplifies what comes next. The wise prepare for the bottleneck they can’t yet see.