THE TOOLS
These tools are the product side of the same operating model: systems built for real constraints, explicit tradeoffs, and outputs a team can actually trust.
HARDSTOP
Deterministic Supply Chain Risk Intelligence
Stop manually checking dozens of feeds. Hardstop monitors your external risk sources, filters the noise, links events to your operational network, and generates auditable daily briefs — entirely offline.
THE PROBLEM
Supply chain risk is not a data problem. You have too much data. The problem is signal — knowing which of the hundreds of daily alerts from NWS, FDA, USCG, and RSS feeds actually touches your network, and which are noise. Most teams either drown in feeds or ignore them entirely. Both are losing positions.
THE DIFFERENTIATOR
Every alert is auditable. Every run has an exit code (healthy / warning / broken). No silent failures. Self-evaluating architecture means the system tells you when something is wrong — not after you've already missed it.
WHO IT'S FOR
Operations leaders, supply chain managers, and risk officers at mid-market companies who need daily situational awareness without a six-figure software contract.
WHAT IT DOES
Monitors public sources: NWS weather alerts, FDA recalls, USCG notices, RSS feeds
Filters noise using configurable suppression rules and trust-tier weighting
Links events to your specific facilities, lanes, and shipments
Assesses risk using deterministic impact scoring — same input, same output, always
Generates tier-aware daily briefs with actionable alerts
All data stays local in SQLite — no cloud, no subscriptions, no data leaving your environment
Genetic programming that evolves alert rules you can read, edit, and hand to the next person — no black boxes, no 'the model said so.'
THE PROBLEM
Most anomaly detection tools hand you a score and say 'trust me.' When an alert fires at 3 AM and your on-call engineer asks why, the answer is usually 'the model said so.' In production systems where people need to understand, debug, and trust their alerts, that is not good enough.
THE DIFFERENTIATOR
Interpretable AI is not an academic concern. In operations, you need to know why something alerted so you can tune it, trust it, and hand it to the next person. GP-evolved symbolic rules give you that transparency without sacrificing the ability to discover non-obvious patterns.
WHO IT'S FOR
Engineering and ops teams running monitoring pipelines who are drowning in alert noise and need a system that discovers non-obvious patterns without sacrificing explainability.
WHAT IT DOES
Breeds populations of explicit symbolic logic rules selected for real-world operational fitness
Enforces precision floors (≥30%), false positive ceilings (≤15%), and sensible alert rate bands
Produces rules you can read: ('if_alert', ('>', ('avg', 'latency'), 100), 'High alert!')
PromotionManager enforces economic constraints — patterns must demonstrate causal contribution or get evicted
Meta-evolution layer breeds its own hyperparameters — no manual tuning
Determinism guarantee: same seed, same Python version, single-threaded = identical results, always
STRAINER-CLI
Offline Document Intelligence for Pipelines
Turn .txt and .md files into structured summaries, tags, and metrics — entirely offline, no API calls, no uploads, no black boxes.
THE PROBLEM
Most summarizers assume a web app, cloud processing, and opaque models. That breaks when notes must stay on disk (privacy, compliance, or operational security), when you need outputs that plug into scripts and agents, or when you need repeatable behavior — not a 'maybe it summarizes' black box.
THE DIFFERENTIATOR
Built for pipelines, not dashboards. Every summary is traceable back to the source — evidence anchors tell you exactly where each claim came from. No network required. No data leaves the machine.
WHO IT'S FOR
Knowledge workers, analysts, and teams managing large document repositories who need structured, machine-readable summaries without sending documents to third-party services.
WHAT IT DOES
Produces extractive summaries — top-ranked sentences from the source, not hallucinated rewrites
Infers topic tags from the text for search, indexing, or retrieval systems
Outputs word count and compression ratio metrics
Evidence anchors map every summary sentence and tag back to its exact position in the source
Output is clean JSON or Obsidian-compatible Markdown — drops cleanly into scripts and agents
Fast mode requires zero dependencies; Smart mode uses local sentence embeddings for higher quality
THE TEMPORAL GRADIENT
Salience-Aware Compute Scheduling
A dynamics framework that modulates processing rate based on information salience — so your system spends compute where it matters, not uniformly across everything.
THE PROBLEM
Most systems treat all inputs equally — processing everything at the same rate regardless of urgency or novelty. A CRITICAL SECURITY BREACH DETECTED alert and a routine status update should not consume the same processing resources or trigger the same response latency. Uniform polling is computationally wasteful and operationally blind.
THE DIFFERENTIATOR
This is a dynamics framework, not a cognitive model. All claims are limited to defined state variables and testable invariants. Designed to be embedded in larger systems — agents, pipelines, monitoring infrastructure — as a principled compute scheduling layer.
WHO IT'S FOR
AI engineers and systems architects building agents or pipelines that need principled, salience-aware resource allocation rather than uniform polling or naive priority queues.
WHAT IT DOES
Maintains an internal timebase (τ) whose rate is modulated by salience load (Ψ)
Salience is a function of novelty and value — high-salience inputs accelerate the internal clock
Memory strength decays over internal time and reconsolidates on access
Configurable cooldown policies prevent rapid repeated reinforcement
Full structured telemetry with canonical schema validation
Deterministic behavior: testable invariants, reproducible state, no hidden stochasticity
OPERATOR TOOLKIT
9 Templates for Managers Who Run Real Operations
A ready-to-use bundle of operations templates for daily management, shift handoffs, weekly reviews, 1:1s, action tracking, issue visibility, KPI review, and staffing stability.
THE PROBLEM
Most managers are running on memory, scattered sheets, messy handoffs, and weekly reviews that produce more heat than follow-through. The result: dropped actions, unclear ownership, repeated problems, and reactive leadership that stays vague until it becomes a fire.
THE DIFFERENTIATOR
This is not a giant operating system. It is a simple working layer for managers who are tired of rebuilding the same tools every week. Start with the spine — Daily Checklist, Shift Handoff, Action Tracker — and layer in the rest as needed.
WHO IT'S FOR
Warehouse managers, operations managers, site leaders, assistant general managers, and team leaders who need more structure without more bureaucracy.
WHAT IT DOES
Daily Leader Checklist — start every shift with clear visibility
Shift Handoff Template — hand off reality cleanly, not from memory
Weekly Operations Review — review patterns instead of just reacting to them
Manager 1:1 Template — coach people with more clarity and less vagueness
Action Tracker — track commitments across time without losing them
Issue & Escalation Log — spot recurring problems before they spread
KPI Review Sheet — measure what matters without dashboard bloat
Staffing Stability Review — see staffing pressure before it becomes a fire
Start Here guide + License and Use Notes included
TOPOGUARD
Topology-Gated Refusal for Dynamical Systems
A calibration-driven control layer that learns the shape of a stable regime, then decides when to automate, intervene, or refuse.
THE PROBLEM
Most monitoring and control layers rely on amplitude thresholds, hand-tuned rules, or black-box anomaly scores. That breaks when a system changes shape without obviously changing magnitude, or when teams need a refusal boundary they can justify instead of another vague confidence score.
THE DIFFERENTIATOR
TopoGuard is built around refusal as an engineering primitive. Instead of pretending every input deserves an answer, it learns what admissible structure looks like and says no when the topology leaves that envelope. That makes the boundary legible, calibratable, and operationally defensible.
WHO IT'S FOR
AI infrastructure teams, controls engineers, and safety-minded builders who need an interpretable gate in front of automation when regime shifts matter more than raw amplitude.
WHAT IT DOES
Calibrates on a known stable regime and classifies future windows as AUTOMATE, INTERVENE/FLARE, or REFUSE/VOID
Measures connectedness, recurrence, and window-to-window structural change instead of relying only on variance
Supports scalar, multivariate, and streaming inputs through one interface
Offers a lightweight proxy backend plus a persistent-homology backend for exact topological analysis
Uses calibration-derived quantiles rather than fixed magic numbers
Keeps streaming behavior bounded and replayable with explicit dimensionality checks
OPERATOR TO OPERATOR
If one of these tools maps to a live operating problem, that is usually where the engagement starts.
The open-source layer proves how the system thinks. The consulting layer adapts that logic to your environment, constraints, and decision structure.
THESE TOOLS ARE HOW I WORK
These are the instruments I use in consulting engagements — when I recommend a solution, I've already built and run it. HARDSTOP is consulting-ready today; Operator Toolkit is available as a paid template bundle; the remaining software instruments are open and in active validation. I describe each product's maturity and availability honestly rather than overselling it. This is the engineering proof behind the Operator Audit: I do not just identify operational drift; I build instruments that make the system easier to see, trust, and run. If any of these could solve a problem in your organization, let's talk.