Research

The mechanics of autonomous execution. Notes from building agents that move value without giving up control.

Selected research. Simulation-first studies, published with the receipts to re-run them.

Simulation

Deterministic simulation before execution

Previewing routes, fees, and outcomes on forked chain state before anything signs.

Compiling boundaries into runtime gates

How written policies become checks the signing edge cannot skip.

Multi-chain settlement under constraints

Route selection when fees, time, and depth all bind at once.

Building and deploying autonomous agents

From typed intents to a governed run loop you drive with the Nex CLI.

Tracing and inference for agent runs

Making every model call and route decision legible in the receipts trail.

x402 monetization for agent services

Charging per request over HTTP 402 with settlement bounded by policy.

Areas

How agents act under boundaries you define, and why every action ships with the evidence that it stayed inside them.

Deterministic preflight on forked chain state — routes, fees, slippage, and failure paths surfaced before anything signs.

Smart order routing across venues and bridges, ranked by policy-adjusted cost, depth, and time-to-settle.

Stablecoin sends, claimable links, and per-request settlement that stay inside a wallet policy end to end.

Method

Every finding begins at a deterministic simulation on forked state, so it can be re-run, not just cited.

Intent, policy decision, simulation, approval, and settled result travel together as one reviewable record.

We cluster failures and near-misses from settled runs and turn them into tighter policies in the open.

Follow the work

New research, simulation write-ups, and post-mortems land on the Nexis blog as they ship.