Operational systems. Deterministic tooling. Research with teeth.
I build tools for operators, researchers, and people trying to think more clearly under pressure.
> whoami
Justin Shank — operator, researcher, builder of deterministic things
> build --philosophy
determinism over vibes
traceability over black-box output
local-first control over forced cloud dependence
systems that fail loudly, not silently
> open --best-work
Hardstop / Dragons & Distractions / The Temporal Gradient / GRBToolkit
| I'm here for... | Start here |
|---|---|
| 🏭 Operations — risk, execution, auditable systems | Hardstop · Alert-Axolotl-Evo · Shank Strategy Ops |
| 🧠 Neurodivergence — ADHD tools, knowledge, humane interfaces | Dragons & Distractions · Cognitive Glow · Psi-Loop |
| 🔬 Research — cognitive architecture, AI safety, signal processing | The Temporal Gradient · TopoGuard · GRBToolkit |
| 🌀 Artifacts — intentional experiments and liminal things | Zero Terminal · Thresholds |
| ✍️ Writing — AI, complexity, the systems underneath | The Edge Explored → |
These aren't constraints imposed by clients or frameworks. They're load-bearing.
- Deterministic — given the same inputs, same outputs. Always. No hidden heuristics.
- Auditable — if a tool fires an alert or makes a decision, you can trace exactly why. Supply chains collapse when nobody knows.
- Local-first — your data lives on your machine. No subscription, no vendor to go dark, no surprises at 3am.
- Loudly broken — tools should fail with exit codes, not silently succeed at nothing.
A few projects that best represent how I think and build:
Hardstop · Dragons & Distractions · The Temporal Gradient · GRBToolkit · TopoGuard
Deterministic supply-chain risk engine for explainable alerts and operational monitoring.
Why it matters: turns external events (news, government data) into traceable, reason-coded risk decisions linked to your internal network — facilities, shipments, lanes.
Best for: ops leaders, supply chain, resilience teams
deterministic local-first auditable production-shaped
Genetic programming for anomaly detection. Evolves explicit logic trees under economic constraints.
Why it matters: fitness metrics align with your actual cost of false positives — not a generic accuracy score. Self-evaluation routines return exit codes for CI pipelines.
Best for: ops automation, anomaly detection, interpretable ML
deterministic auditable interpretable production-shaped
AI brand-studio workspace combining market analysis, generative design, and supply-chain monitoring.
React 19 + Tailwind frontend, Express/Supabase backend. PDF reports, offline-capable, integrated tests.
active prototype
Monster-themed ADHD support. Printable ritual cards, metadata scripts, and a companion card preview repo.
Why it matters: humane tools for people whose brains don't organize like spreadsheets.
production-shaped local-first
Obsidian plugin. Tracks note openings, computes a glow score (recency + frequency), surfaces top notes in a sidebar.
Why it matters: your best ideas are usually buried — this makes them float.
Tunable decay rates, focus mode. active prototype
Retrieval ranking library that prefers useful, non-redundant context. Ranks candidates by value relative to goal × novelty relative to what you already have.
Pluggable embedders, benchmark scripts vs. similarity-only baselines. research notebook
Offline summarisation for .txt and .md files. Key sentences, tags, metrics, evidence anchors.
Fast (keyword) or Smart (embedding) mode. No server, no API key. local-first production-shaped
Browser-local checklist with categories, moods, priorities, quotes, backup/restore, bulk undo.
Zero server dependency. local-first
Inserts currently playing track into Obsidian notes. Daily history, custom templates.
Offline packaging and deterministic dependency management for the Obsidian plugin/theme registry.
Research on time-weighted attention and memory consolidation for AI agents.
Why it matters: explores what it means for an agent to have a genuine sense of recency — not a flat context window but a signal that decays meaningfully.
Interactive experiments on signal decay, retrieval order, forgetting curves. research notebook
Formal verification for prompt injection detection. Models state transitions topologically, proves safety properties.
Why it matters: "we tested it and it seemed fine" is not a safety argument. active prototype
Working paper on information-theoretic causality for AI systems. Proofs, definitions, open questions. research notebook
Modular analysis suite for gamma-ray burst light curves.
Supports synthetic models (QPIX, FRED), real Fermi data, Bayesian block sweeps, Wavelet Z (WWZ) computation, AIC model comparison.
Why it matters: reproducibility-first — scripts produce CSVs and figures, notebooks capture full analyses, citation guidelines included.
deterministic research notebook
FastAPI/SQLAlchemy reference implementation for deterministic biological data ingestion, semantic resolution, and lineage-backed trust workflows.
Every ingestion step logged. Every canonical record traceable. REST routes for ingestion, sources, canonical records, semantic assertions, review, lineage, workflows.
deterministic auditable production-shaped
Not every project here is a product. Some are probes, interfaces, or constructed environments.
A terminal-themed personal website where you navigate by typing commands. Virtual file system, CRT effects, 30 commands mapping to posts, projects, and life statuses. Vanilla JS + Vite, deployable on Vercel.
A digital doorway disguised as a repository. Portal page, generative sigils, a manifest of states, hidden base64 keys. For people who like liminal spaces and cryptic YAML. The mood is as intentional as the code.
Open-source mod menu for Gorilla Tag (VR). Forked to champion community modding and keep the GPL chain intact.
Official site for Shank Strategy Ops LLC. React 19 + Vite, Express backend, Radix UI, Tailwind, serverless contact forms.
Personal site. See Zero Terminal for the theme.
On the surface this portfolio spans supply chains, ADHD tooling, astrophysics, and strange little experiments. Underneath, the same structure keeps appearing:
| Stage | What it solves | Repos |
|---|---|---|
| Signal | Detect what matters from noise | Alert-Axolotl-Evo · GRBToolkit · Hardstop |
| Selection | Choose what's worth attention | Psi-Loop · Cognitive Glow · Strainer-CLI |
| Decision | Make a traceable, auditable call | Hardstop · TopoGuard · Biological-Data-OS |
| Interface | Present it to a human legibly | Lumina-Pro · Zero Terminal · Journey-Log |
| Adherence | Help humans actually act on it | Dragons & Distractions · The Temporal Gradient |
The research layer (Information-Causality, Temporal Gradient) is about making sure each stage is theoretically grounded, not just empirically functional.
The Edge Explored — essays on AI, complexity, and the systems underneath.
From the Familiar — a series written from inside the work: what it looks like when a mind that doesn't sleep collaborates with one that can't stop thinking.
I run Shank Strategy Ops — execution leadership for complex projects across supply chain, knowledge management, and AI safety.
If you need deterministic tools that won't gaslight your auditors, or execution leadership on something that keeps slipping — get in touch or type contact in Zero Terminal.
May your data be structured, your alerts be meaningful, and your dragons be friendly.



