Reid Hoffman: The contrarian bet that built LinkedIn

Hoffman looks for founders who are “insanely ambitious” yet realistic about odds and competition. When he started LinkedIn, he told early employees they had only a 20–25% chance of success, then worked obsessively to improve that probability. He expects entrepreneurs to understand their competitive landscape deeply, avoid saying “we have no competitors,” and articulate why they can be first in ‘Glengarry Glen Ross’ style markets where winner-takes-most outcomes are common.

He argues you don’t get to an optimistic future by trying to avoid failure. The key is taking *smart* risks: going into areas where smart people think you’re wrong, but you have a coherent theory of why they’re mistaken. LinkedIn and Airbnb both started as ideas that many smart people dismissed as crazy. Hoffman framed the perceived risks—e.g., people afraid LinkedIn signaled disloyalty, or cities and hotels hating Airbnb—as precisely why they had years of low competition to establish network effects.

Hoffman believes if a founder isn’t spending roughly a third of their time on hiring, they’re underperforming. He prioritizes references over interviews, prefers people with steep learning curves over long CVs, and loves the Zuckerberg heuristic: hire people you’d be willing to work for. He stresses that entrepreneurship is a team sport—long-term success requires leaders who attract, grow, and retain A-players, not just those with a good product idea.

He’s blunt that “work–life balance is not the startup game.” Early LinkedIn worked six and a half days a week; parents were expected to go home for dinner and then come back online afterwards. He argues this intensity is voluntary but non-negotiable if you want to compete globally. Founders should be honest in recruiting—clearly setting expectations about long hours and intensity—rather than selling a fantasy of easy balance.

Hoffman frames AI as a general-purpose technology like electricity or the printing press: it will cause disruption and some harm, but the long-run upside is enormous. He thinks current models are essentially cognitive copilots or “idiot savants” that already have superhuman breadth (e.g., comparing mixture-of-experts AI architectures, game theory, and oceanography—something no human can do). His advice: individuals should immediately start using AI for real work, using role prompts (e.g., “act as my critic,” “act as a historian of technology”) to expand their thinking and skills.