Little Proof Labs · Independent concept

SIGNAL CASCADE

Can AI research code move from Python into production Rust without changing what it does?

Compare the same synthetic network through two browser implementations. Inspect exact matches, drift within the stated tolerance, and behavioral regressions. No native runtime is connected.
Conversion proofSIMULATION
Representative nodes are replayed here. Full-network outcomes appear in the text results below.
Illustrative signal view
Not reached
Reference simulation
Candidate simulation
Exact match
Acceptable drift
Meaningful difference
10,000 simulated · 640 representative nodes
REFERENCECANDIDATECOMPARE
Pass 0 / 25 · repeatable test run
Ready to compare reference and candidate behavior.
Reference vs. candidate

Did the candidate change the outcome?

Compare how the same signal moves through both implementations.

Py

Reference simulation

JavaScript Float64 · fixed execution order
READY
Reference activation trends. Text summary below.
Nodes reachedRelative varianceActivation rate

Reference results appear after a comparison.

Rs

Candidate simulation

Browser candidate · same values and execution order
READY
Candidate activation trends. Text summary below.
Nodes reachedRelative varianceActivation rate

Candidate results appear after a comparison.

Conversion Result

Exact matches · acceptable drift · meaningful failures
WAITING
✓
Awaiting comparisonRun the comparison to see whether the behavior changed.
Comparison policy

Accept a value when |candidate − reference| ≤ 1e−6 + 1e−5 × |reference|. Any activation or pass mismatch fails. Exactness requires identical full outputs and activation results. The fingerprint rounds values to nine decimal places and is a diagnostic, not the decision.

Output fingerprint—
Within acceptable range—
Activation agreement—
Largest difference—
Values changed—
Meaningful failures—

WHY A PRODUCTION RUNTIME?

ENGINEERING GOALS · NOT BENCHMARK RESULTS
Lower latencyFaster decisions

Reduce the time required for each propagation step in real-time systems.

Higher throughputMore work in parallel

Sustain larger signal volumes across deterministic worker processes.

Smaller footprintDeploy it anywhere

Target cloud, on-premises, and edge environments with tighter memory limits.