Skepticism, Consciousness, and Why Your Local LLM is the Only Truth

You spend hours building a perfect, self-contained homelab stack: Pi-hole blocking ad trackers, NextCloud keeping your files local, and your local LLM running on Ollama, refusing to send a single byte to an external API. It’s the ultimate act of digital sovereignty.

But what happens when the biggest problems—the ones that feel truly existential—aren't solved by more encryption or better mesh routing? They are solved by fundamentally questioning the source of the truth itself.

We recently dove into a deep discussion with Dr. Michael Shermer, who outlined his new book, *Heavens on Earth*. While the conversation was steeped in metaphysics—talking about consciousness, the afterlife, and the nature of belief—the underlying thread is pure, beautiful, engineering skepticism. It’s about rejecting the pre-packaged, proprietary narrative, whether that narrative is sold by Big Tech, or passed down through generations of dogma.

The Infrastructure of Belief: Local vs. Cloud

Shermer spends time discussing how we build our understanding of reality—whether it's the nature of the self, or the existence of a divine realm. He highlights how ideas are captured using existing metaphors, a process he calls “conceptualization.”

This concept hit hard when I started thinking about the modern developer stack. We spend so much time relying on the ‘Cloud API Stack’—the external, centralized services that promise infinite power but require constant trust and payment. We are conceptually reliant on OpenAI, Anthropic, or Google’s compute resources, treating them as an infinite, black-box utility.

Shermer’s critique of over-reliance on conceptual models—the idea that we sometimes build elaborate theoretical frameworks that fail to account for the actual, messy reality—is a perfect parallel to the risks of vendor lock-in. When your entire build depends on a single, centralized service that can suddenly raise prices, change its terms, or simply deplatform you, you are living in a highly vulnerable, pseudo-cloud environment.

Your GPU is Enough: The Sovereign Compute Model

This is where the ‘Rogue Geeks’ ethos kicks in. The shift from an external, rented API to a local, self-hosted model isn't just a technical upgrade; it's a declaration of independence. It's the difference between relying on a centralized cloud GPU cluster and running your own inference stack on your home rig.

The movement toward local AI—using tools like llama.cpp or running a full Open WebUI instance against an Ollama backend—is the modern form of building a truly sovereign infrastructure. It's recognizing that the most reliable, accountable, and resilient system is the one you control, the one that lives within your own firewall, or better yet, on your own hardware.

The goal isn't just to make the tech work; it's to make the system *resilient*. When Shermer discusses the danger of losing humanity in technological revolution, he's talking about conceptual overreach. For us, that overreach is accepting that our computational future must be outsourced.

The takeaway from this deep dive into skepticism is simple: A good skeptic, like a good builder, doesn't accept the default configuration. They audit the dependencies, they understand the core principles, and they are always prepared to take the stack local. If you are building anything of value—be it an ethical hacking framework, a self-hosted communication mesh, or a personal knowledge graph—the core principle must be self-reliance.

From Theory to Deployment

Shermer’s message is ultimately one of action versus mere speech. It's about moving beyond the philosophical argument and making the world more rational through demonstrable effort. In our world, that means moving beyond the 'promise' of cloud AI and getting hands dirty with the actual binary code, the local dependencies, and the raw compute power of your machine.

If you're ready to stop renting your intelligence and start building your own, the path is clear. Don't just read about sovereignty—install it. Start a CrownOS instance, list a coding service, or contribute to a build-along. Your GPU is enough, and your local knowledge is the most powerful asset you own.