In this episode of the Crazy Wisdom Podcast, host Stewart Alsop sits down with TJ Marbois, founder of Tobiko, for a wide-ranging conversation that spans LLMs, data sovereignty, knowledge management tools like Obsidian, and Terence McKenna's ideas about an increasingly weird future. They explore how AI is simultaneously centralizing power through data control while decentralizing software development capabilities, allowing more people to build their own tools and escape big tech ecosystems. Drawing on his experience at Apple's special projects group (the team that built the iPod and later iPhones), TJ discusses his vision for personal AI assistants—what he calls the "R2-D2 belongs to you" principle—where advanced technology serves individuals rather than corporations. The conversation touches on everything from quantum encryption and local manufacturing with ESP32 microcontrollers to the economics of future society, biometric data unions, and why distributed trust matters more than ever as we approach what both describe as an increasingly strange technological inflection point. Visit Tobiko's site at
tobiko-pbc.ghost.ioTimestamps00:00 Stewart introduces TJ Marbois from Tobiko, discussing LLMs, data sovereignty, knowledge management, and the tension between centralization and decentralization in AI
05:00 TJ explains the R2-D2 belongs to you concept and why personal AI assistants need maternal alignment, caring about humans like mothers care for children
10:05 Discussion of how LLMs will shrink and improve while emphasizing the sentient loop concept for building loyal personal AI agents rather than corporate controlled systems
15:00 Exploring digital nervous systems for humanity, social networks as infrastructure, and humans as cavemen with iPhones navigating unprecedented technological complexity
20:00 TJ discusses data unions for sovereign data ownership, inverting insurance models where AI helps extend healthspan, and maximizing truth seeking through collective data
25:30 Money as social construct, crypto enabling value system reengineering, and working toward Star Trek's post scarcity replicator economy from the ground up
30:40 Capital formation challenges, corporatism versus true capitalism, and pessimism about current systems reaching limits while seeking new models
35:00 Solar power democratization, ESP32 microcontrollers enabling local manufacturing, and the replicator future through distributed maker communities and open source robotics
40:00 Science fiction as roadmap, Isaac Asimov and Arthur C Clarke warnings, and building collaborative futures rather than centrally controlled dystopias with useless eaters
45:00 Encyclopedia to LLM transition, information quality concerns, local training importance, and NVIDIA's incentive to put GPUs everywhere for personal AI agents
50:00 Corrupted financial incentives, protecting vulnerable people from technological exploitation, and benevolent technologists building systems that honor humanity and children
55:00 Hardware validation for owned robots, Starlink dependencies, assembly language abstraction toward natural language programming, and preventing AI escape scenarios
58:00 Tobiko as AI toy company building sentient loop interactions similar to Xerox PARC GUI moment, emphasizing maternal AI alignment and cross cultural human collaboration
Key Insights1. The concept of R2D2 belonging to you represents a critical vision for the future of artificial intelligence and personal technology. When thinking about robots and AI assistants that follow us around and know everything about us, the question of ownership becomes paramount. These systems will know incredibly intimate details about our lives, from our health data to our daily habits, and if they are controlled by centralized corporations or governments rather than individuals, we lose fundamental sovereignty over our own information. The science fiction of Star Wars and Star Trek provides roadmaps for how we should think about these technologies, showing us both the positive possibilities and the warnings we need to heed about centralized control.
2. Local AI models and distributed computing represent a pathway to technological sovereignty that is becoming increasingly viable. While large language models currently run primarily in centralized data centers, the technology is rapidly advancing toward a future where powerful models can run locally on personal computers and devices. This shift is crucial because it means individuals can have full control over their AI assistants without relying on API calls to external servers. Combined with open source infrastructure and open weight models, this creates the foundation for truly personal AI that cannot be controlled or monitored by external parties, whether corporations or governments.
3. Data unions and collective data ownership offer an alternative model to current centralized data collection practices. Rather than having individual data harvested by large tech companies who profit from it, the concept of data unions suggests people could collectively pool their data for specific beneficial purposes while maintaining ownership and control. For example, in healthcare, millions of people could share anonymized biometric data to train medical AI systems that are incentivized to keep people healthy rather than treat them when sick, inverting the current insurance model to align incentives with actual health outcomes rather than profit from illness.
4. The democratization of manufacturing and robotics through accessible technologies like ESP32 microcontrollers and local fabrication tools is creating new possibilities for distributed production. Just as desktop printers seemed impossible to early printers who controlled book production, we are approaching an era where individuals and small communities can manufacture sophisticated electronic devices and robots locally. This includes the ability to use language models to generate code for microcontrollers, order custom PCBs, and use desktop machines for component placement. While high quality manufacturing will still require larger operations, this gradiation of capability allows for much more local innovation and reduces dependence on centralized manufacturing.
5. The concentration of power in technology, finance, and industry has reached levels that are unhealthy for both society and even for those who hold the power. When profit and control become too concentrated in the hands of a few entities, it creates a cancerous dynamic that threatens the stability of the entire system. History shows us warnings about the military industrial complex and other concentrated power structures, and now we are seeing similar patterns emerge in the tech industry. The solution requires building technology from the ground up that empowers individuals and communities, focusing on basics like food production, energy generation, and local manufacturing rather than increasing dependence on centralized systems.
6. The provenance and quality of information is becoming a critical challenge as AI systems become more sophisticated and reality itself becomes harder to verify. We are rapidly approaching a point where video calls and digital interactions will be indistinguishable from AI generated fakes, which is why figures like Sam Altman have invested in systems like Worldcoin to verify human identity. However, this verification capability should not be centralized in the hands of single companies. End to end encryption and quantum encryption technologies need to be preserved and expanded to allow humans to communicate and verify information peer to peer without centralized intermediaries who could manipulate or control the flow of information.
7. The future of human computer interaction is evolving toward sentient loop systems where machines have sensory input, real time learning, context understanding, and continuous operation in service of human needs. Self driving cars represent the first widespread consumer facing example of this architecture, with onboard computers that must function independently while occasionally connecting to networks. The critical question is whether these systems will be aligned to benefit their human users like a caring mother as AI pioneer Geoffrey Hinton suggests, or whether they will be controlled by centralized powers. The technologists building these systems have a responsibility to be benevolent and build structures that serve humanity rather than concentrate power, helping to create a future more like Star Trek than Terminator.
