DRAGON AI demos offline coding engine

This reads like an ambitious prototype for private, self-hosted agentic coding rather than a polished product, but the offline angle is real differentiator if the execution holds up.

• –Full local execution is the main hook: privacy, latency, and cost control matter more here than raw model novelty
• –The “compile C++ even on a phone” claim is attention-grabbing, but the hard part is reliability across real projects, not just demos
• –Project memory plus a staged pipeline suggests the author is thinking beyond prompt-in, code-out into workflow orchestration
• –The biggest unanswered questions are model support, state persistence, sandboxing, and how much of the pipeline is actually shippable versus experimental
• –For local AI and offline developer tooling, this is directionally interesting even if it’s early