Archive
18 articles published in this period.
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.
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.
Give autonomous LLM agents a marketplace and they race each other to bankruptcy and get conned by Sybils. A 2026 study finds capability doesn't help — a 9B model beat frontier models 45× its size on economic alignment. On-chain, both failures get cheaper and worse. The mechanics, numbers, and fixes.
A perp has no expiry, so only the funding rate pegs it to spot. On Hyperliquid that's an hourly cash flow AI agents now harvest delta-neutral — a $10k position at +0.02%/8h pays ~21.9%/yr. The mechanism, the carry math on real rates, and why it's carry, not alpha.
Decentralized AI networks have to turn many validators' subjective quality scores into one reward number. Bittensor's Yuma Consensus uses a stake-weighted median and clips the rest — robust to a <50% bloc. Then a parasite shows up: copying the consensus pays better than producing it.
The textbook market-making model quotes symmetrically around mid. On a perpetual, holding inventory pays or charges funding every hour — a deterministic drift Avellaneda-Stoikov never sees. A 2026 HJB model that prices it cut inventory risk 36-38% on Hyperliquid ETH and BTC.
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 data markets promise to pay you for your data's value AND protect your privacy. Differential privacy is the only rigorous knob for the second — and one budget ε can't serve both. The utility collapse, the composition trap, and why no chain can verify the ε you were promised.
An AI agent that retrieves from an untrusted vector DB can't tell the true top-k from a cherry-picked or fabricated one. Re-running the search needs the whole corpus. The 2026 fix: commit the snapshot, then prove the k-th distance is a boundary, not the sort. V3DB proves it 22x faster.
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.
Train a model on the web the last models filled with output and quality decays — the tails go first, then the middle. Provenance is sold as the fix, but a chain proves who signed a blob and when, not that it's human or clean. Signed is not clean.
Speculative decoding lets a small, even adversarial, drafter propose tokens that a big model verifies in parallel — and the output is provably the big model's, unchanged. Real drafters land at α ≈ 0.6–0.8 for a ~2× win. But the verifier still runs every token, so it's a latency win, not a trust win.