Nodes, indexers, RPCs, GPUs, and the operational backbone of AI and chain systems.
Decentralized RL splits the actor from the learner across the internet: the policy a worker acts with runs several steps behind the one that learns. INTELLECT-2 held reward at 4 steps stale; SparrowRL cut the broadcast 79x. The bandwidth and staleness taxes, and what the chain secures.
Verification overhead gets the blame, but the real tax on decentralized LLM inference is the batch a single-tenant node can't fill. On identical 8x H100s, Llama-3.3-70B costs about $258 per million tokens at batch 1 and $2.30 at batch 256 — a 112x spread. Why, and how networks fight it.
Secure multi-party computation runs a transformer without any party seeing your prompt — at minutes per query and gigabytes per token. BumbleBee's BERT-base: 6.4 GB, 2.55 min; LLaMA-7B: ~8 min/token. The bytes don't go where you'd think, and it's why 'blind' networks fall back to TEEs.
On-chain AI loves to say a model is 'stored on-chain.' But a chain commits to a 32-byte hash for cents; keeping the 140 GB it points to retrievable is a separate, recurring, surprisingly centralized bill. The storage math, erasure coding, and the retrieval wall.
A DeFi liquidation mints a fixed prize, and for years it went to whoever won the gas war, not the protocol. Aave's own data shows Chainlink SVR routing $675M of liquidations and clawing ~$16M back. Inside the recapture-auction mechanics — SVR, Oval, API3 — and what they don't fix.
A decentralized GPU network has to answer two questions an AWS invoice never raises: is this the GPU it claims, and did it actually do the work? io.net logged 327,000 registered GPUs and roughly 6,720 daily-verified. We dissect metadata trust, PoW puzzles, performance fingerprinting, and zkGPU-ID.
Passive AMM liquidity is a short option that arbitrage bots exercise every block. LVR = σ²/8 prices the rent — at ETH's 63% realized vol that's ~5% of pool value a year, and most pools don't earn enough fees to cover it. The math, real numbers, and the auctions clawing it back.
Most decentralized-AI networks vote or average. Allora bets the network should learn whom to trust per context — using forecasters that predict who's about to be wrong, mapped to weights through a softplus gradient. We dissect the mechanism, the math, and where it pays.
An autonomous agent that swaps in the open broadcasts its intent to every searcher in the mempool. We solve the optimal sandwich on a live Base pool, show why an agent's slippage default is the searcher's profit knob, and price the defenses.
You pay per token for a named model; the provider picks the precision. FP8 quantization costs 0.6 MMLU points and is near-invisible to output auditing — so inference markets bond and attest instead of detect.
Permissionless training networks pay peers for gradients they can't re-run. Proof-of-Learning was meant to verify the work — until it was forged for 3% of the training cost, then for one floating-point op per weight. Here's the mechanism, the attacks, and what's deployed instead.
DX Terminal Pro ran 3,505 LLM agents trading real ETH on Base for 21 days: 7.5M invocations, ~$20M volume, 99.9% settlement. The reliability came from the operating layer around the model, not the weights — here are the numbers and the failure modes.