On building anti-fragile software products
I’ve been dabbling with the idea of anti-fragility — a concept from Nassim Nicholas Taleb’s work that changed how I look at software. He describes three kinds of systems:
- Fragile systems break under stress. Like glass - they shatter when dropped.
- Robust systems withstand stress. Like stone - solid, stable, unchanged.
- Anti-fragile systems gain from stress. Like fire - when fed, they grow stronger.
That lens stuck with me. What makes a software product fragile, robust, or anti-fragile?
Fragile software: These systems couldn’t adapt. They weren’t bad — just brittle in the long run.
- Flash - Dominated for years, but couldn’t survive the shift to open standards.
- Clippy - Early AI, but limited and too rigid in its usefulness.
- Vine - Perfect in form, but couldn’t evolve past its 6-second constraint. They mattered once. But over time, they cracked under pressure.
Robust software: They stay steady — reliable, built to last as the world shifts.
- PostgreSQL - Reliable and battle-tested.
- Microsoft Office - Still relevant in a changing world.
- Linux - Absorbs change through modularity. These tools last. They survive.
Anti-fragile software: This is where it gets interesting. These systems learn. They grow from chaos.
- Cursor - AI-native code editor that improves with use.
- ChatGPT - Trained on noise, thrives on ambiguity.
- Midjourney - The more it’s used, the more expressive it becomes. • The cloud - Built to fail, recover, scale — and get better.
Anti-fragile products aren’t just stable — they evolve under pressure.
This mindset shift is huge.
Founders, PMs, engineers — we need to ask:
“Is what I’m building fragile, robust, or anti-fragile?” Revenue and traction are good signals — but they don’t tell you how your system handles stress.
Do your products learn? Adapt? Get better as they grow? That’s the real test.
I’m still dabbling with this idea. Looking at it mostly from a product lens, with a metaphorical take on the tech side.