Journal
136 articles on the craft of AI × blockchain engineering. or filter by topic:
Ethereum's blob fee market held at 1 wei for twenty months after Dencun, then broke in December 2025. Here's what the data shows, what drove the cycle, and what it means for L2s and AI agents that depend on cheap calldata.
On-chain AI agents reset state on every call. Four persistent-memory architectures span a 20× monthly cost gap — SSTORE at $78/month versus IPFS+CID at $3.92. Real June 2026 gas numbers, four trust models, one crossover you won't see coming.
To slash a dishonest AI operator on EigenLayer you need a proof. To produce one, you need determinism. EigenAI solves both: seeded inference + optimistic re-execution inside a TEE collapses verification to a byte-equality check — slashable at Ethereum-scale security.
Speculative decoding accelerates LLM inference 2–4× by splitting work between a cheap draft model and an expensive verifier. When those run on different nodes in a decentralized network, three properties of the protocol that are trivially safe on one machine become open attack surfaces.
PREVRANDAO costs roughly $88 per bit to bias. Chainlink VRF costs $2.43 per request. Here's how to choose the right randomness source for AI agents making on-chain decisions — and what breaks when you choose wrong.
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.
An SP1 proof costs 327k gas on Ethereum mainnet. At a 30M-gas block target, a naive AI inference market maxes out at 91 verifications per block. SnarkPack's O(log N) aggregation collapses 1,000 proofs to a single 880k-gas verification — here's the mechanism, the math, and the latency tradeoffs.
DiFR's seed-commitment round-trip is the last synchronization barrier to async verifiable inference. TOPLOC removes it: a locality-sensitive hash of intermediate activations compresses to 258 bytes per 32 tokens — 1,000× smaller than raw embeddings — and validates faster than inference ran.
A 15× spike on a $17M pool shifts its 30-minute v2 TWAP by 9.3% in one block. The arithmetic-mean accumulator is far weaker than v3's geometric tick accumulator. Covers the manipulation cost formula, the Inverse Finance case study, and minimum safe pool depth.
AI agents now resolve most prediction markets straight off the web — UMA's bot hits 99.3% on sports and 72% on mention markets. The optimistic oracle's dispute game is the backstop, but it only catches errors someone is paid to catch. The numbers, and the silent-settlement gap.
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.
Every zkSNARK bottoms out in a pairing check. Before Pectra you had BN254 at ~100-bit security. EIP-2537 added BLS12-381 at 128 bits, and its pairings are cheaper: 32,600k+37,700 vs 34,000k+45,000 gas. G2 MUL dropped from millions to 22,500 gas.