A $200 Claude subscription can deliver almost $10,000 worth of usage, but Jonathan Archer thinks those economics will not last.
Jonathan Archer is co-founder and CEO of Open LLM, a gateway designed to put multiple AI models, subscriptions, and API keys behind one layer. Before Open LLM, he spent roughly six years advising crypto projects across business development, growth, and community. He met his co-founder at the Avalanche Summit in Buenos Aires before the pair began building side projects together.
Open LLM started with a simple frustration: hitting AI subscription rate limits and then wasting time maintaining open-source alternatives. Jonathan’s thesis is that the future will be multi-model, not dominated by one model or one interface. Different models already have different strengths, and users should be able to combine them without constantly moving between products.
That leads to his more provocative argument: AI companies should not trap users inside proprietary harnesses. “It’s their tokens,” Jonathan says. Open LLM is being built so users can bring subscriptions and API keys, maintain a fallback chain across providers, share inference between devices, and eventually treat multiple machines as part of one AI infrastructure layer.
The conversation also moves beyond routing. Jonathan describes a future where agents spin up machines and subagents, pull inference from another authenticated device, and use lower-cost models for smaller jobs rather than burning the most capable model on everything. Open LLM is also exploring unified memory, one-click VPS deployment, multi-device sessions, and enterprise routing.
In Today’s Episode We Discuss
- 00:01Jonathan Archer and Open LLM
- 03:34The Origin of Open LLM
- 05:30AI Subscription Economics
- 07:28Subscription Routing and Compliance
- 09:01Unified AI Memory
- 11:04Multi-Device AI Infrastructure
- 13:19Self-Hosted Models and VPS Deployment
- 14:43Open LLM Pricing and Free Tier
- 16:19Cross-Model Workflows
- 17:03The Technical Challenge of AI Routing
- 19:10The Multi-Model AI Future
- 20:24Free vs Paid Plans
- 21:37Fallback Chains and Model Selection
- 23:38AI Subagents and Token Efficiency
- 24:35Open Source vs Closed Infrastructure
- 26:12Multi-Device Sessions and Enterprise AI
- 28:15Breaking AI Platform Lock-In
- 29:53OpenRouter and the Gateway Model
- 32:06AI Gateway Acquisition Value
- 34:09Open LLM and Vercel
- 36:18Dynamic Model Discovery
- 37:38Beyond the AI Gateway
- 38:29Rapid-Fire Questions
- 39:00Crypto Adoption and Real-World Utility
- 40:56Crypto, Payments, and Stablecoins
One of the most concrete techniques is Open LLM’s configurable fallback chain, which moves to another model when usage runs out or a provider goes down. Jonathan also explains why subagents should default to less capable models for smaller tasks, comparing the alternative to sending Steve Jobs to retrieve a file. And he describes dynamic discovery that continuously queries providers for newly available models.
The bigger idea is older than AI: durable infrastructure often wins by preserving choice rather than forcing everything into one closed system.
Pull Quotes
“It’s their tokens, they should be able to go.”
“We’re gonna live in a multi-model future.”
Follow Jonathan Archer on LinkedIn: LinkedIn
Visit Our Website: Website
🎧 Watch, listen, and follow on your favorite platform: Website
📝 Read the full episode breakdown on our blog: Website