OAPN #4: Pay the People

OpenAgents AfterParty at TEMS: Energy, Bitcoin Mining, and AI Compute

Live from OpenAgents Headquarters in Austin, Texas Car talks with guests at the TEMS afterparty [https://luma.com/wzaw7joe] after The Energy and Mining Summit. We talk with guests about energy, Bitcoin mining, AI compute, Austin’s builder scene, and where the next wave of infrastructure is headed. The night continued from the summit to the OpenAgents Studio, with open conversations and high-signal afterparty energy. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit openagents.substack.com [https://openagents.substack.com?utm_medium=podcast&utm_campaign=CTA_1]

OAPN #8: Ocean Power: Oceanic Phase Three

Ocean Power: Oceanic Phase Three | OAPN #8 In OAPN #8, we preview the oceanic Phase Three of the OpenAgents compute network in conversation with Ben Silone. This episode explores what ocean-based infrastructure could mean for the future of compute, energy, and machine work. We discuss what we admire and question about Panthalassa’s floating datacenter vision, revisit the Bitcoin Magazine thesis that bitcoin can help unlock ocean energy for large populations, and look at OceanBit’s efforts to combine bitcoin mining with ocean thermal energy. We also cover this week’s decentralized training run through Pylon v0.2, including a legal-focused Qwen fine-tune and how that fits into the OpenAgents flywheel of revenue and rev-sharing. Along the way, we reflect on the AI labs racing to build centralized systems for governments and institutions, versus the OpenAgents effort to distribute capability outward and put more power in more hands. Links:Website: https://openagents.com [https://openagents.com]Docs: https://docs.openagents.com [https://docs.openagents.com]GitHub: https://github.com/OpenAgentsInc/openagents [https://github.com/OpenAgentsInc/openagents]Psionic: https://github.com/OpenAgentsInc/psionic [https://github.com/OpenAgentsInc/psionic]X: https://x.com/OpenAgentsInc [https://x.com/OpenAgentsInc] This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit openagents.substack.com [https://openagents.substack.com?utm_medium=podcast&utm_campaign=CTA_1]

OAPN #7: Worse is Better

OAPN #7: Worse is Better In this episode, recorded from the new OpenAgents headquarters in downtown Austin, we discuss the two competing design philosophies of software: the New Jersey school and the MIT/Stanford school, and how those philosophies map onto the AI landscape today. We get into why the “Worse is Better” tradition matters, why simple systems with strong distribution often beat more elegant closed systems, and why Bell Labs, C, and Unix still offer some of the best lessons for anyone trying to build durable infrastructure. From there, we connect that philosophy directly to OpenAgents. Rather than disappearing for a year behind closed doors to build some secret superintelligence, OpenAgents is starting from something much simpler: can we create products people actually want to use, can we price resources the market currently values at zero, and can we begin a real economic flywheel around compute, agents, and bitcoin? We talk through the basic insight behind the OpenAgents compute market: there is an enormous pool of stranded consumer compute in the world, but without buyers its effective price is zero. If OpenAgents can create demand for that compute and produce something economically useful from it, the result is not just a product but the beginning of a new market. It will not be perfect. It will not have the budgets of the big labs. But if it is simple, useful, open, and capable of paying people, it has the ingredients for something that can grow. Links: Website: https://openagents.com [https://openagents.com] Docs: https://docs.openagents.com [https://docs.openagents.com] GitHub: https://github.com/OpenAgentsInc/openagents [https://github.com/OpenAgentsInc/openagents] X: https://x.com/OpenAgentsInc [https://x.com/OpenAgentsInc] #OpenAgents #OAPN #Bitcoin #AI #MachineWork #OpenSource This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit openagents.substack.com [https://openagents.substack.com?utm_medium=podcast&utm_campaign=CTA_1]

OAPN #6: Developer Bounties

Developer Bounties In this episode, we relaunch the OpenAgents bitcoin bounty program, paying developers for high-quality additions across our codebases. We also walk through the broader product roadmap and explain how OpenAgents is thinking about the next phase of the stack. The goal is ambitious and clear: build the OpenAgents product suite into a real open alternative to what the big labs offer, especially Anthropic. That means building a product experience that can increasingly serve as a drop-in replacement, powered by open models, open infrastructure, our growing compute network, and eventually the models we train ourselves, supplemented where useful by frontier systems. We discuss what that roadmap looks like, where contributors fit into it, and why bounties matter as a way to bring more builders directly into the work. If you want to help build open agent infrastructure and get paid in bitcoin for doing it, this episode is for you. Links: Website: https://openagents.com [https://openagents.com] Docs: https://docs.openagents.com [https://docs.openagents.com] GitHub: https://github.com/OpenAgentsInc/openagents [https://github.com/OpenAgentsInc/openagents] Psionic: https://github.com/OpenAgentsInc/psionic [https://github.com/OpenAgentsInc/psionic] X: https://x.com/OpenAgentsInc [https://x.com/OpenAgentsInc] #OpenAgents #Bitcoin #DeveloperBounties #OpenSource #AI #MachineWork This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit openagents.substack.com [https://openagents.substack.com?utm_medium=podcast&utm_campaign=CTA_1]

OAPN #5: Distributed Training 101

In OAPN #5, we introduce the basics of distributed language model training. Using educational material from Nathan Lambert’s RLHF Course and Stanford’s CS336, we walk through the foundations of conventional model training and explain the core concepts in a way that connects directly to what OpenAgents is building. We then introduce DiLoCo, or Distributed Low-Communication Training, and discuss how projects like Bittensor, Templar, and Prime Intellect have approached distributed training, along with how the Psionic and Pylon approach differs. This episode is meant to help people understand what is actually happening under the hood when their machines participate in networked training and why that matters for the future of open AI infrastructure. If you want to understand what runs on your computer to earn bitcoin as part of a distributed training network, this is the place to start. Links: Website: https://openagents.com Docs: https://docs.openagents.com GitHub: https://github.com/OpenAgentsInc/openagents Psionic: https://github.com/OpenAgentsInc/psionic X: https://x.com/OpenAgentsInc This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit openagents.substack.com [https://openagents.substack.com?utm_medium=podcast&utm_campaign=CTA_1]