Archive
142 artifacts, grouped by date.
Decode speed of a frontier MoE model versus interconnect bandwidth. Each layer's expert routing is an all-to-all shuffle bounded by the slowest link. Drag from NVLink down to home broadband: the line crosses the interactive floor at the datacenter edge, then falls to seconds per token.
A perpetual market maker's quote ladder. Avellaneda-Stoikov skews only for inventory; funding adds a second, signed skew that survives even at zero inventory. Drag the reservation line to set inventory, push funding and horizon, and watch the two skews stack or cancel.
Seven ways to verify what an AI agent claims it did, plotted by latency added against the share of fabricated claims caught. Pick a claim type and watch it route to the cheapest sufficient verifier — a free chain receipt, a 12ms signed receipt, or a zkLLM proof reserved for the inference itself.
On-chain cost of a query versus how much history it touches. Re-executing pays a cold SLOAD per value and hits the 60M block-gas wall — or can't reach the data at all. A coprocessor's proof verifies flat at ~300k gas whatever the size. Drag the cursor or pick a real query; the cyan line never moves.
Cumulative net carry of a $10k delta-neutral perpetual funding harvest over real Hyperliquid funding regimes. The line starts below zero by the round-trip fee, then climbs at the annualized funding rate; a flip slider rolls it over. Carry, not alpha — until funding inverts.
An on-chain agent is told to pay an invoice, but one MCP tool hides an instruction in its description that redirects the transfer to an attacker. Pick a defense layer — model, tool, client, policy, custody — and run it; alignment alone refused under 3% of these in MCPTox.
An inspection game between a solver and a verifier. Drag the operating rate and tune the bond, reward, and bounty to find p*, the minimum sampling rate that makes honest AI inference a dominant strategy — and watch it collapse without a bounty.
Bittensor's Yuma Consensus scores a miner by the stake-weighted median of validator weights, then clips everything above it. Validators sit on a cumulative-stake axis; the consensus line is read at the κ=50% mark. Push a colluding bloc past 50% and watch the pump go from clipped to unstoppable.
In decentralized RL the rollout worker acts with a policy several steps behind the trainer. Staleness g = broadcast time / step cadence. Pick a model and link, toggle sparse deltas, and watch g cross INTELLECT-2's demonstrated 4-step budget — full 32B weights blow right past it.
Cost per million tokens and throughput for Llama 3.3 70B on 8x H100 as the batch fills. A single-tenant node (B=1) runs ~25 tok/s at ~$258/M; a filled batch (B=256) runs ~2,800 tok/s at ~$2.30/M — same silicon, 112x cheaper. Pick a batch; read the tax you pay for the GPU you can't fill.
Under secure two-party inference, a transformer's cost is eaten by Softmax and GELU, not the matrix multiplies. Stacked per-operation communication and interaction rounds for one BERT-base block across Iron, BumbleBee, and Nimbus — the nonlinear activations are 80–95% of the bill.
Proving an agent's top-k retrieval is real costs a sort over every scanned candidate, unless you prove a boundary instead. Drive top-k and the corpus preset; watch the in-circuit comparison count split into the flat boundary proof and the sort tax you skip, anchored to V3DB's measured 22x.