a16z's joshelman: harness + context can lock in users even without model moats

joshelman · x · 2026-09-08

Replying to a debate on AI lock-in, a16z partner @joshelman argues that even if agent training loops become the moat, the abundance of similarly capable models advancing together means harness + context with network effects can be its own lock-in — a counterpoint to @ruslansv's claim that only major labs can build model-level moats.

Related event: VCs Debate AI Moats: Training-Coupled Harnesses vs. Generic Ones(4 posts)→

Original post →

More from AGI Musings

AGI Musings channel →