Litepaper

Useful compute, honestly decentralized.

Cloudana is a decentralized compute network where the work that secures a provider's reward is the user's computation.

Independent providers contribute CPUs, GPUs, and storage. Users run real workloads — AI/ML, scientific computing, web hosting, batch jobs — and pay in CLD through smart contracts on Base. For matrix-heavy workloads, Cloudana's Proof of Useful Work makes the cryptographic proof and the useful result the same operation. No wasted puzzle energy. No "just trust the provider."

How it works

Flash and go. A provider installs once. The node auto-detects its CPU, RAM, disk, and GPU, classifies its tier, and self-registers. No spec forms. No bidding marketplace.

You pick what to run. Want to run an AI workload? Select the model and scale you need. We classify it, route it to a node that fits, and handle placement. The heterogeneity is hidden behind one choice.

Proof = product. For matrix workloads, the provider proves the work by computing it. The proof's hardness comes from the full transcript of the computation — unfakeable — and the decoded result is the answer you paid for. Verification is mathematical, not re-execution.

The token

Built to last, not to cap out

Most chains either inflate forever or halt emission entirely. A useful-work network can't halt — the work must keep happening — so CLD has no hard cap. Instead:

Disinflationary emission with a permanent tail. Starts at 8%/yr, decays 15%/yr, never below 1.5%. Providers are always paid to keep the network alive.
2% of every workload payment is burned. Because users pay real value for real compute, CLD scarcity grows with usage — not speculation.
A soft cap set by the market. When usage is high, burn outpaces emission and supply becomes deflationary. When low, the tail keeps the lights on. The network self-regulates instead of relying on a hardcoded number.

CLD already has value because every workload is someone paying for work.

Where we are — said plainly

Cloudana's testnet is trust-minimized and orchestrator-coordinated, not yet "completely decentralized." Verification is re-run by a trusted coordinator today. Full trustlessness arrives when on-chain zkSNARK verification removes the coordinator from the critical path — that's our mainnet milestone, and we'll claim "decentralized verification" when it's true, not before.

Differentiators

Proof of Useful Work for AI/ML and scientific compute — proof and product coincide.
Demand-backed token — emission tied to real work, scarcity tied to real usage.
Plug-and-play onboarding — hardware detection, not configuration.
Multi-tier verification — PoUW where it fits, challengers where it doesn't.

Run a provider node Read the full whitepaper