Open Source AI Business Models and Brand Moats

Blog post, October 3, 2023. By Michael Dempsey (Eads Bridge Holdings).

Key Takeaways

• The traditional open-source model breaks under AI economics. Building a managed service on top of a public good assumes the good is cheap to maintain; high capex and rapid model obsolescence break that.
• Models go stale and get swapped out as inference layers modularize, so competing at the model layer is economically unsustainable for most labs.
• Brand moats beat incremental performance. Developer loyalty follows “personality or an opinionated aesthetic” rather than benchmark wins.
• Crypto is the analogy — mercenary capital rotates toward brand identity and community trust, not raw technical advantage.
• The personification window has closed. Altman leveraged personal brand at OpenAI; a startup now trying to occupy “AI thought leader” is wasting effort. Niche community differentiation is what is left.
• Likely endgame is duopolistic, mirroring mobile: one dominant open platform, one closed alternative, with real value capture confined to specialized domains.

Connections

• Saved against Bubble Architecture, and it is the structural counterpart to Clem Delangue’s prediction from the same week: both agree the model layer commoditizes, and disagree about who is left holding anything afterward.
• The “all platforms start as features” argument in Modularization of Software is the same claim aimed at company strategy rather than at moats.