The Threat of the Cloud
For marginalized communities—especially those in conflict zones like the India-Myanmar border—Cloud-based AI represents an existential threat. It requires high-speed internet that is often embargoed by the state, and it extracts indigenous data into foreign corporate servers without compensation. Furthermore, it rarely supports local dialects, accelerating cultural erasure. In this session, we will open the floor to discuss these systemic threats and share our collective experiences.

The Challenge: Designing Intelligence at the Edge
Rather than a traditional presentation, this will be an interactive workshop. We invite developers, activists, and thinkers to brainstorm architectures for the "Digital Commons." We will discuss the realities and challenges of running small-parameter models (3B to 7B parameters) using Llama.cpp on accessible edge hardware, like a Raspberry Pi. How do we best utilize quantization to fit these tools into 8GB of RAM? How can we power these systems entirely by 12V DC solar setups in the jungle? We want your insights.

Retrieval-Augmented Generation (RAG) as a Survival Tool
Finally, we will workshop ways to prevent AI "hallucinations" in critical survival scenarios. We will collectively brainstorm how to connect a local LLM to offline vector databases of agricultural PDFs and medical manuals. By designing pipelines that synthesize answers strictly from locally curated texts, we aim to map out a future where a farmer can safely diagnose crop diseases locally—without a single packet of data ever leaving the village. Bring your ideas, your questions, and your expertise as we approach this challenge together.