1meg
Back to Labs
Active Research

Distributed Computing

Moving computation from centralized clouds to the edge and beyond. WebAssembly enables portable code; decentralized networks enable portable infrastructure.

The Compute Spectrum

Centralized Cloud
50-200ms

Traditional data centers (AWS, Azure, GCP). High latency, single points of failure, concentrated control.

Examples: AWS EC2, Azure VMs, GCP Compute

Edge Cloud
10-50ms

Servers at the edge of networks, closer to users. Major cloud providers and CDNs.

Examples: Cloudflare Workers, AWS Lambda@Edge, Fastly Compute

Decentralized Compute
Variable

P2P networks where anyone can provide compute. No single provider controls the network.

Examples: Akash, Golem, Flux, iExec

Local/On-Device
<1ms

Computation on the user's own device. Zero network latency, complete privacy.

Examples: WASM in browser, local AI models, native apps

Technologies

WebAssembly (WASM)
Production

Portable binary format that runs at near-native speed. Write once in any language, run anywhere - browser, server, edge, embedded.

Learn More
WASI
Maturing

WebAssembly System Interface. Standardized API for WASM to interact with the OS - files, network, clock. Enables portable server-side WASM.

Learn More
Cloudflare Workers
Production

JavaScript/WASM execution at 300+ edge locations. Cold starts under 5ms, scales to millions of requests.

Learn More
Akash Network
Production

Decentralized cloud marketplace. List your spare compute, rent others'. 85% cheaper than AWS for some workloads.

Learn More
Golem
Production

P2P marketplace for computing power. Focus on batch processing - rendering, ML training, scientific computing.

Learn More
Fly.io
Production

Deploy apps globally with automatic placement near users. Full VMs, not just functions.

Learn More

Key Concepts

Edge Computing

Processing data near where it's generated rather than sending to centralized data centers. Reduces latency, bandwidth, and privacy exposure.

Serverless

Code runs in response to events without managing servers. Provider handles scaling, you pay per execution. But still centralized.

Decentralized Compute

P2P networks where anyone can be a provider. No single entity controls the infrastructure. Censorship-resistant by design.

Portable Execution

WASM enables code to run identically across different environments - your laptop, a server, a browser, an embedded device.

Use Cases

Use CaseDescriptionBest Layer
Real-time GamingGame logic at the edge reduces latency to <20msEdge
AI InferenceRun models in browser or at edge, no data leaves deviceLocal/Edge
Video ProcessingTranscode, filter, analyze video streams near originEdge
Batch Rendering3D rendering across thousands of nodesDecentralized
Scientific SimulationClimate models, protein folding across P2P networkDecentralized
Privacy-Preserving ComputeProcess sensitive data on user's deviceLocal
The Vision: Compute Everywhere

Today, "the cloud" means a few massive data centers owned by three companies. Your data travels hundreds of miles to be processed, then back again.

The future is different: computation happens wherever it makes sense. Latency-sensitive code runs at the edge. Privacy-sensitive code runs on your device. Batch jobs run on whoever has spare capacity.

WebAssembly makes code portable. Decentralized networks make infrastructure fungible. The result: computing as a utility, like electricity - you don't care which power plant generated it.

Beyond the Cloud

We're researching distributed compute architectures that don't depend on centralized providers.