Replicate is one of the most popular AI inference providers, reporting over 2 million users as of their $40m Series B with a16z. But how did they get there?

Ben Firshman, CEO of Replicate, came on the pod to talk about why ML researchers are bad users, going through YC during COVID, how AI art communities influenced their product roadmap, and where Replicate is going next.

Full show notes: https://latent.space/p/replicate

Timestamps:

00:00:00 Introductions
00:01:22 Low latency is all you need
00:04:39 Evolution of CLIs
00:06:47 How building ArxivVanity led to Replicate
00:13:13 Making ML research replicable with containers
00:19:47 Doing YC in 2020 and pivoting to tools for COVID
00:23:11 Launching the first version of Replicate
00:29:26 Embracing the generative image community
00:31:58 Getting reverse engineered into an API product
00:35:54 Growing to 2 million users
00:39:37 Indie vs Enterprise customers
00:42:58 How customers uses Replicate
00:44:30 Learnings from Docker that went into Cog
00:52:24 Creating AI standards
00:57:49 Replicate's compute availability
01:02:38 Fixing GPU waste
01:10:58 What's open source AI?
01:15:19 Building for AI engineers
01:17:33 Hiring at Replicate