Pharma does 800 small bets, why can’t software?

I just read a report about Eli Lilly’s “clock speed” approach. Basically, they break drug development into ~800 tiny decisions. Most fail, but the few that work get scaled big, and that’s how they keep producing blockbusters. It made me wonder why software doesn’t do the same. Instead of big roadmaps, why not run hundreds of small bets, kill the weak ones fast, double down on the strong ones. Software should be even easier to test than drugs, but in reality we keep building huge projects that flop. Has anyone here actually seen this approach used in dev teams or startups?