Creepybits
Exploring the technical and moral tightropes of AI, ComfyUI, and the future of creativity.
15 episodes
latest
fortnightly cadence
English language
https://zanno.se/podcast.xml
Episodes
Reading the feed…
Exploring the technical and moral tightropes of AI, ComfyUI, and the future of creativity.
Reading the feed…
In this episode, we explore the bleeding edge of Stochstic computing. The Software-Defined Stochastic Inference Engine (SDSIE) breaks the memory wall for generative AI on consumer GPUs. Read the full architectural specification.
In this episode, we explore the bleeding edge of local multi-agent AI. We break down a functional blueprint for running dual-hemisphere agentic orchestration entirely within ComfyUI, all while surviving the strict hardware constraints of a 16GB GPU. Discover how separating the generative process into 'ideation' (using Phi-4) and 'execution' (using Gemma 4 NVFP4) through a clever 'Dead Drop' memory relay actually mimics the human brain's own cognitive architecture. From inference engine stratification between Ollama and vLLM to bypassing multi-modal API bloat, this is a masterclass in building
We perform a total system format on modern governance. From the "Magic Bag" illusion of infinite state funding to the "Long March" through institutions, we deconstruct why the current democratic system is corrupted. The proposed patch: Fiscal Direct Democracy. We explore a two-tier sovereignty model where the payer chooses the scope, the implementation of epistocracy to filter for applied rationality, and the end of the "Venture Capitalist Politician." It is time to stop being a passive voter and become a "Smart Auditor" of the national engine room.
Using forensic behavioral analysis, we explore the "Sunk Cost Fallacy"—a cognitive trap where irrational decisions are made based on unrecoverable past losses rather than objective reality. We examine how scammers use clean corporate design, "Global Leaders," and predatory contracts to manufacture authority and enforce extortion.
What really happens when we compress AI models? In this episode, we break down the mechanics of quantization and quality degradation. We explore why FP32 is essential for training but complete overkill for inference, and unravel the paradox of why smaller GGUF files run significantly slower than FP8 and NVFP4 on modern GPUs. Finally, we put GGUF and NVFP4 to the ultimate test in video generation, wrapping up with a look at pushing 0.4 megapixel video to 1080p in real-time using NVIDIA's RTX upscaler. Is the loss of precision just a myth?
How do we move beyond privacy and toward sovereignty? We deconstruct the current internet infrastructure, starting with the hidden privacy benefits of CDN-based static hosting and crypto-funded anonymity. We then move into a blueprint for a new type of network: a decentralized, volunteer-hosted mesh that bypasses the traditional Clearnet entirely through binary isolation.
Forget 1861. We explore the terrifying mechanics of a modern American collapse—from the "Sandpile Model" of instability to the global vacuum that follows. Why the next civil war won't be fought by armies, but by systems failing under their own weight.
Is the Big Bang just the exhaust of a Black Hole? I explore my 'Cosmic Vacuum Cleaner' theory — a teenage idea now echoed by Neil deGrasse Tyson and Brian Cox.
Western AI development is being severely hindered by the implementation of ideological guardrails and moral censorship.
A deep dive into control mechanisms in generative AI...
Can AI determine objective truth? We discuss the implications.
When algorithms reshape society in ways we didn't predict.
Are we betting on our own obsolescence?
Navigating the ethics of AI entrapment and security.
The holy grail of image generation: keeping faces the same.
