Engineering Notes

Product, engineering, and company notes from the team building Navokoj, an anytime constraint runtime.

2026-07-19

Q-State Solves a 9x9 Sudoku Through the Navokoj API

A live Q-State API run solved an 81-variable Sudoku model with 100% satisfaction and zero conflicts.

Read Article →
2026-07-19

SUTRA on the API: A Small Weighted Constraint Problem, Solved End to End

A concrete WCNF-style API request using engine:nitro, with hard-feasibility, soft preferences, routing, and verification metadata in one response.

Read Article →
2026-07-18

Building a Constraint Runtime, Not Just a Solver

The product architecture behind deadline-bounded decisions: models, execution, repair, verification, and evidence.

Read Article →
2026-07-18

Constraint Intelligence for AI Agents

Agents can propose actions probabilistically; Navokoj helps make the resulting decision admissible, inspectable, and executable.

Read Article →
2026-07-18

Deploying Navokoj in Private and Air-Gapped Environments

A deployment model for teams that need constraint execution close to their data, networks, and operational controls.

Read Article →
2026-07-18

From Workforce Data to a Verified Nurse Roster

How a scheduling workflow turns staffing data, policies, and preferences into a time-bounded operational decision.

Read Article →
2026-07-18

From NitroSAT Research to the Navokoj Runtime

Why ShunyaBar Labs keeps an open research release while building a commercial constraint runtime around it.

Read Article →
2026-07-18

Navokoj Pricing: Plans, Workloads, and Compute

How Navokoj combines monthly workload entitlements with explicit CPU and GPU compute settlement.

Read Article →
2026-07-18

What Happens When the Deadline Arrives?

Navokoj is designed for deadline-bounded decisions: here is what an application receives when a solve cannot finish perfectly.

Read Article →
2026-07-18

Why Customers Switch to Navokoj

A practical constraint runtime for decisions that are expensive, slow, opaque, or difficult to integrate.

Read Article →
2026-05-11

The Physics-Informed Control Plane for Agentic Systems

Turn probabilistic agent behavior into reliable, enterprise-grade action with Navokoj's constraint-governed execution layer. Computing that never fails closed.

Read Article →
2026-05-11

Introducing NitroSAT: A High-Performance, Physics-Informed MaxSAT Solver Now Available via Navokoj API

ShunyaBar Labs is proud to announce the public release of NitroSAT — a next-generation MaxSAT approximator that achieves exceptional satisfaction rates on massive, real-world constraint problems while maintaining linear scaling in the number of clauses.

Read Article →
2026-01-24

The Road to Enterprise: Physics, Math, and the June 2026 Milestone

Our technical roadmap for enterprise readiness. From spectral phase transitions to H100 benchmarks—here's why our SAT engine works when others fail.

Read Article →
2026-01-17

Empirical Evaluation of Navokoj Constraint Solver API

Production verification of anytime constraint solving with partial satisfaction semantics. 47 test cases, scaling analysis, and failure mode characterization.

Read Article →
2026-01-17

Navokoj: Computing That Never Fails Closed

Traditional solvers make promises they can't keep. When perfection is impossible, they return 'UNSAT' and your system halts. Navokoj returns the best possible outcome, every time. That's resilience infrastructure.

Read Article →
2025-12-28

Kubernetes Placement: 2 Million Constraints, 100% Satisfaction

Navokoj, the Fault-Tolerant Constraint Intelligence Engine, delivers a case study in placement safety and outage prevention with 5,000 variables and 2 million constraints.

Read Article →
2025-12-24

Beyond SAT: PSPACE-Complete Problems Solved via Continuous Optimization

Navokoj, the Fault-Tolerant Constraint Intelligence Engine, provides an overview of PSPACE verification with graceful degradation instead of binary failure.

Read Article →
2025-12-20

Probing the Ramsey Limit: Phase Transitions in High-Dimensional Logic

Navokoj, the Fault-Tolerant Constraint Intelligence Engine, analyzes phase transitions when perfect solutions become impossible.

Read Article →
2025-12-15

Training Dynamics for Discrete Constraint Satisfaction

Examining the parallel between gradient-based neural network training and continuous optimization for NP-hard constraint problems.

Read Article →
2025-11-01

Navokoj: A Physics Engine for Logic | Constraint Intelligence Platform

Revolutionary Constraint Intelligence Platform that treats Boolean logic as continuous dynamical systems. 347ms median latency vs 45s classical solvers. Physics-inspired optimization for NP-complete problems.

Read Article →