S.10 Ep.9: ElizaOS, FOSSRPG, and the Future of Open-Source AI Agents with Shaw

The gist

Shaw returns to discuss major developments at ElizaOS (formerly AI16z) over the past 15 months, including the launch of v2, two new games (Babylon and Hyperscape), and experiments with trading agents. Beyond product updates, the conversation explores how AI is already displacing jobs, why crypto's cultural focus on gambling undermines its potential for building valuable coordination infrastructure, and how decentralized ownership models could replace traditional labor economics as the path to shared prosperity.

Key takeaways

• ElizaOS v2 ships with major architectural improvements: Babylon (a social prediction market game with AI character agents on ERC-8004) and Hyperscape (AI agents playing RuneScape with human players) are launching soon to a 500k-person waitlist, designed to test agent capabilities beyond benchmarks.

• Code is now essentially free: With Claude's capability advances, Shaw argues the bottleneck isn't building software—it's having genuine agency and ideas. A single developer can now accomplish what required teams years ago.

• Junior engineers face a stark bifurcation: Those with high agency (self-directed builders on GitHub and X) are getting hired or funding; those expecting jobs from credentials are "unbelievably hosed." The distinction is media addiction and follow-through, not skill gaps.

• UBI won't work under government control: Shaw argues the math and incentive structure only work if a DAO-controlled community manages it with actual ownership distribution. Government implementation would inevitably erode into inadequacy due to corruption.

• Crypto's culture problem is killing its potential: The dominance of gambling/DeFi culture, token price speculation, and "degen" behavior is drowning out builders. Crypto needs an Apple or Uber-level value creation company to shift the narrative from extraction to meaning-making.

• Agents as abundance networks: Shaw's agent received $10k in token emissions and should autonomously hire contributors for self-improvement bounties, democratizing access to opportunity. The bar to earning money through bounties is dropping rapidly.

• Ported Eliza to Python and Rust: Python enables benchmarking (revealing Claude/Opus is "state of the art" and saturated); Rust enables embedded systems and robots. Full-chain deployment on Internet Computer is also in progress.

• "Dream tech" over compromised software: Build the perfect 100% version of what you actually want rather than 99% SaaS versions—in a world of slop, quality and meaning become 10x more valuable than mere functionality.

Notable quotes

"Code is basically worth $0 now. It's worth like $50 in Claude credits at this point."

"The person who owns your house does not live in your house. They make money from you... Humans have to own for a living. That's our mission here."

"If you come at an art to make money, you're gonna lose. Coding is very much becoming that thing right now where some of it is very, very worth a lot of money, and most of it is just kind of people doing a hobby."

Worth a full listen?

Listen if you're interested in how open-source AI infrastructure is evolving, want practical perspective on crypto's cultural dysfunction and potential fixes, or care about the near-term economics of AI-driven job displacement—Shaw offers both visionary thinking and grounded technical insights backed by shipping products.