Superlinear Returns
Paul Graham’s essay Superlinear Returns explains why outcomes in the real world are so unequal—and why this matters for how we approach our work.
The Core Insight
We’re taught to think linearly: twice the effort yields twice the results. But in many domains, performance returns are fundamentally superlinear. A product half as good as competitors’ doesn’t capture half the market—it captures none.
This explains the extreme inequality in outcomes we see everywhere: a few papers get most citations, a few startups capture most value, a few researchers make most discoveries.
Two Sources of Superlinear Returns
1. Exponential Growth
When success breeds more success, outcomes compound dramatically. Learning is a prime example—knowledge builds on knowledge. Once you understand the fundamentals, each new concept becomes easier to acquire.
2. Thresholds (Winner-Take-All)
Step functions create discrete outcomes. In academic publishing, being above the acceptance threshold matters enormously; being just below yields nothing. The difference between acceptance and rejection can be marginal, but the outcomes are binary.
These sources often intertwine—crossing thresholds enables exponential growth, and rapid growth helps overcome thresholds.
Implications for Research
This framework changes how I think about research strategy:
Seek compounding domains. Work in areas where knowledge accumulates and builds on itself. My work on LLM multi-agent systems benefits from this—each project deepens understanding that accelerates the next.
Cross thresholds early. Getting your first paper accepted, your first grant funded, or your first collaboration established creates momentum. The first is hardest; subsequent ones benefit from reputation and experience.
Follow curiosity over prestige. Graham notes that fields with superlinear returns typically require independent-mindedness. Original ideas matter more than execution in these domains.
The Changing Landscape
Graham observes that technology has expanded opportunities for individual achievement. This creates both greater potential rewards and greater risk—higher variation in outcomes benefits the exceptional but hurts the average.
For researchers, this suggests doubling down on what makes you distinctive rather than following the crowd. Safe, incremental work in a superlinear world may be the riskiest strategy of all.
Practical Heuristics
Graham offers several useful heuristics:
- Always be learning continuously
- Seek fields where a few outliers vastly outperform others
- Notice gaps at knowledge frontiers
- “Do things that don’t scale” initially
The last point is counterintuitive but important. Early in a research career, it’s worth doing labor-intensive things that don’t scale—manual data collection, detailed case studies, intensive collaborations—because they help you cross initial thresholds.
My Reflection
Understanding superlinear returns has made me more patient with slow early progress and more aggressive about pursuing compounding opportunities. The graph of research impact isn’t linear—it’s exponential with step functions. The question is whether you’re building toward those inflection points.
Where do you see superlinear returns in your field? Reach out at persdre@gmail.com.