Context
Flare's Rust-to-WASM compilation makes it ideal for edge runtimes. Fermyon Spin achieves sub-0.5ms cold starts. WASM 3.0 (W3C Standard, Sept 2025) standardizes threads + SIMD.
Proposal
- Create `flare-edge` crate: minimal WASM component for serverless inference
- Support WASI-nn proposal for standardized ML inference
- Target Cloudflare Workers, Fermyon Spin, Deno Deploy, Fastly Compute
- Sub-millisecond cold start, no container overhead
- OpenAI-compatible API endpoint in a WASM module
Impact
- Deploy LLM inference to 200+ edge locations with zero infrastructure
- Pay-per-request pricing, scale to zero
Context
Flare's Rust-to-WASM compilation makes it ideal for edge runtimes. Fermyon Spin achieves sub-0.5ms cold starts. WASM 3.0 (W3C Standard, Sept 2025) standardizes threads + SIMD.
Proposal
Impact