Innovation velocity
The numbers behind the pace.
Every metric on this page traces back to a test, a seed, and a verdict in a live evidence register. The platform that produces this evidence is itself one of our products.
The scale of the evidence
SolvSRK was engineered from over 1 trillion scalar-dimension solver observations, over 1.0 billion validated data points (the nearest competitor has 30,000), across 3.5 million instrumented trial runs — 821 billion function evaluations, each computing derivatives across multi-dimensional ODE systems, recorded and mined for precision, stability, and cost signatures.
Timeline
Findings, products, and milestones in order.
A short slice of the record. The full register sits behind audit controls; any claim on this site can be traced to a test ID on request.
Cadence
Findings registered per week during the build sprint
From a cold start to over 1 trillion observations and 1 billion validated data points, one engineer.
- Finding
Noise-survival benchmarks (SSH-1 through SSH-4) published internally
Validated 91%+ solver survival across additive, multiplicative, impulsive, and non-stationary noise regimes where standard solvers survive at 0%.
- Finding
Hybrid solver cross-system precision advantage confirmed
Precision gains of 4 to 8 orders of magnitude on three distinct ODE system classes, permutation-tested at p_adj < 0.002.
- Milestone
Dead-zone transparency built into the product
Where competitor solvers genuinely win, the comparison table says so. Honest boundaries are now a first-class feature, not a footnote.
- Product
SolvBench product requirements frozen
Desktop application architecture agreed: SolvBench control plane, SolvScout characterizer, SolvTune benchmarker, SolvEngine runtime, encrypted profile database.
- Finding
Bounded Network Enrichment validated on 590,540 real transactions
16% reduction in false declines, 23% increase in fraud caught, simultaneously, on IEEE-CIS data with five-fold cross-validation.
- Product
Research Lab reframed as standalone product category
The AI-orchestrated ResearchOps platform that validated the solver has produced 3.5M+ trials, 1B+ validated data points, and 821B+ function evaluations with zero retracted findings. Commercialization path defined.
- Milestone
SolvSRK licensing model published
Two-tier model (Sim and Embed), per-program embedded licensing that matches how defense primes already budget for middleware.
Supporting numbers
Beneath the headlines.
Each test is pre-registered with hypothesis, statistical threshold, and dead-zone boundary. Each trial carries a seed and a timestamp. Each finding carries a verdict from an independent phase-audit agent. Each claim is registered against a specific finding. The audit trail is the product.
What’s next
The roadmap.
Directional, deliberately high-level. What we are building next is either already in the evidence pipeline or in an active board-approved research track.
Free tier with the full characterization and benchmarking experience, Pro tier unlocking the Solv solver family.
Same solver binary deployed edge and cloud, with the data assimilation layer collapsed inside the integrator.
Calibrated uncertainty envelopes from a single solve, packaged for certification-heavy industries.
Bounded enrichment as a pattern extends beyond card authorization. Next: insurance underwriting and clinical triage.