What is Ambient?

Ambient is a Solana‑compatible, proof‑of‑work Layer 1 that turns decentralized GPU compute into one continuously improving, open‑weights foundation model—then verifies the model’s outputs. It delivers the privacy, censorship‑resistance, and cost profile of Crypto AI with the speed, uptime, and reliability developers expect.

Why Ambient

Closed AI is powerful but risky. Centralized providers can change behavior, raise prices, restrict access, censor, or compete with you.
Current crypto‑AI networks underdeliver. Proof‑of‑stake model “marketplaces” have low utilization, stale models, inconsistent quality, and limited incentives for new compute to join.

How Ambient works

Proof of Logits (PoL): verified inference as work
- During generation, the model’s internal logits (pre‑softmax scores) form a unique “fingerprint” of its “state of thinking.”
- Miners commit hashed progress markers of these logits; validators can verify a response by recomputing just one token and checking the hash—expensive to generate, cheap to validate.
Continuous PoL + leader election
- The network continuously measures validated work from inference, fine‑tuning, and training.
- Leaders are elected based on recent logit stake (LStake)—contribution to useful work—while transaction ordering remains Solana‑fast and non‑blocking.
Query auction & QoS
- Users/agents post requests with desired latency and price; miners bid and are refunded upon timely completion. Validators are chosen with LStake‑weighted randomness.
Sharded training & inference
- The network applies sparsity‑aware, fault‑tolerant sharding (inspired by PETALS and D‑SLIDE) so very large models are tractable across heterogeneous nodes—backing the “single, huge model” design.
Privacy & data access
- From client‑side PII obfuscation up to future FHE, plus HTTP/BitTorrent oracles for large datasets—designed for anonymous, censorship‑resistant use at scale.