What does a price leave out?
A simulated quote is an estimate. Its standard error tells you how much precision your computation actually buys.
+ AN OPEN WINDOW INTO ON-CHAIN FINANCE
Watch a vortex take shape. Then trace the connection between moving markets, delayed observations, and the models that describe them.
Fluid mechanics × on-chain observation
Integrating the flow field…
Manual reference · no market observation is implied.
Streamlines integrated through a Burgers velocity field. Blue filaments spiral toward the axis; copper filaments start within the core. Changing Re stretches the display vertically and narrows it radially while preserving volume. The shape illustrates a statistic; it does not simulate market dynamics.
READING THE FIGURE
The fluid mechanics describes the picture. A separate market measurement controls its stretch.
A negative radial velocity moves fluid toward the centre. The inward paths become tighter as their radius decreases.
Above and below the middle, fluid moves away from the centre plane. Radial compression balances axial extension.
Circulation wraps the paths around the axis. Viscosity keeps the rotational speed finite near the centre.
FROM A FLUID TO A PRICE FEED
The original research compares a market’s price-change speed with an empirical scale fitted from oracle updates. A larger ratio means the market is moving faster relative to that fitted scale. It does not guarantee an oracle error, liquidation, or a future price move.
Live mode reads a fixed USDC/WETH Uniswap v3 pool at two Ethereum blocks 25 blocks apart, using their actual timestamps. Re = |Pnow / Pbefore − 1| × 100 / elapsed minutes / κ. The ETH calibration κ = 0.08551457790019718 percent per minute comes from the original study. USDC is assumed equal to USD; pool basis or manipulation can affect this reference. The Chainlink answer is shown separately and its age is reported.
The LINK demonstration is a synthetic nine-second pulse reaching Re = 26.4, a rounded peak reported for 22 August 2026. It is not a replay of historical time-series data. Re = 1 and 5 are descriptive reference values, not universal safety thresholds.
The figure integrates 142 paths with fourth-order Runge–Kutta, using a = 0.4, Γ = 3.6 and core radius 0.36 in dimensionless coordinates. Its display mapping is λ = 1 + 1.9 min(1, (Re / 26.4)^0.55); axial scale is λ and radial scale is 1/√λ. Their volume multiplier is exactly 1.
Original study ↗Original observability lab ↗Inspect our calculation ↗
01Understand the mechanism
02Experiment with the assumptions
03Follow the evidence
FIG. 02 / EXPERIMENT NOTEBOOK
Explore sampling uncertainty and observation delay with two reproducible experiments. These two models use synthetic data; the live Ethereum reading belongs to Figure 01 above.
Average-price call · $100 starting price · 1 session
THE RESEARCH NOTEBOOK
Start with the question.
Then go as deep as you like.
A simulated quote is an estimate. Its standard error tells you how much precision your computation actually buys.
On-chain simulation has a budget. Reusing calculations makes room for more paths and smaller sampling errors.
Markets remember the size of a move better than its direction. That changes the way an option model sees risk.
The market moves. The feed catches up. Explore what happens in the gap between a measurement and its arrival.
Independent reading notes on the St0kes research series. Original authorship and source links are preserved in every brief.
BUILT TO BE QUESTIONED
Every useful model has boundaries.
Ours should be easy to find.
The experiment runs in your browser. A fixed seed and identical inputs reproduce the same paths and results.
Standard error measures simulation noise. It does not cover model error, stale data, or the risk of a real position.
Research findings link to their original source. Historical associations stay distinct from predictions and live measurements.
CURIOSITY IS THE STARTING POINT