The usual vocabulary has two positions. Fragile things break under stress. Robust things survive it unchanged.

Nassim Taleb's observation in 2012 was that this list is missing an entry. Some things get better under stress — muscles under load, immune systems under exposure, a skill under difficulty. The word antifragile was coined because the language had no term and the absence was making the category hard to see.

The underlying shape

Strip the vocabulary and the claim is about the shape of a payoff, which is why it is checkable rather than merely evocative.

Something is antifragile when the upside from a large move exceeds the downside from an equally large move. A position with limited loss and unlimited gain benefits from volatility, because bigger swings are worth more to it. That property is convexity and it long predates the term — it is what an option is.

So the practical question is never whether a thing feels robust. It is: if things get much more variable, does that help me or hurt me? A business whose worst month is survivable and whose best month is uncapped wants variability. One with a fixed upside and an unbounded downside does not, whatever it feels like day to day.

Where it applies honestly

Small experiments with bounded cost. Trying ten things where each costs little and one might work very well. The downside is capped at the cost; the upside is not. Optionality of this kind is genuinely convex, and it is why optionality is worth paying for.

A varied client base. More than robustness, if losing a client reliably teaches you something you apply to the rest. The learning is what makes it convex rather than merely diversified.

Skills acquired under difficulty. Work that stretches you produces capability that easy work does not, which is the whole case for deliberate practice.

Reputation built through visible failure handled well. Rare, real, and not something to arrange deliberately.

Where the term is used loosely

Most things called antifragile are robust, and the difference matters because the strategies diverge. Robustness is bought by adding margin — reserves, redundancy, slack. Convexity is bought by changing the shape of the exposure — capping the loss, uncapping the gain. Adding reserves to a position with unlimited downside does not make it antifragile; it makes it last longer.

The barbell is the applied version: put most of the position somewhere very safe and a small part somewhere with uncapped upside, and hold nothing in the middle. For a business that is a stable core of recurring work plus a small deliberate allocation to things that could be large, and no medium-risk bets that carry real downside without real upside.

The honest limit

The framework is better at describing positions than at generating them, and most business exposures cannot be made convex by choosing to be. You cannot make a client relationship antifragile by deciding to.

What it does reliably is force a question worth asking about any commitment: what is my worst case, what is my best case, and are they the same size? Where they are, variability is a coin flip. Where the best is much larger than the worst, more variability is good and stability is what costs you.