Internal engine, external impact
Wanatux handles the invisible work: scraping, data flows, schedulers, AI endpoints, and operational automations. The value appears through client-facing brands that sit on top of this stack.
Epistemic Instrumentation Lab · Agentic Systems Research
The engine room behind everything else.
Wanatux Labs designs and runs automation, data, and AI systems—and studies how models activate, ignore, and misuse knowledge over time. Quiet, opinionated, built for shipping—not for show.
Alpha Release
Sigil is now publicly available
The first public Sigil alpha is live as a research preview. It is small on purpose: local-first, read-only, and synthetic-data-backed so the core memory pattern can be inspected without exposing private infrastructure.
About the lab
What Wanatux is for
Wanatux Labs is not a public agency or SaaS front. It is the long-running personal forge where infrastructure, automation, and experiments are shaped into tools that other brands can safely use.
Opinionated by design
The lab is built around a few principles: ship small, keep things local when possible, and use AI as an amplifier—not as a crutch. Each tool should earn its place in the stack.
Surface of the work
What Wanatux builds
Most of the lab stays offline by design. The list below captures the kinds of systems and tools that live here and quietly support day-to-day operations.
Automation & Ops
Schedulers & back-office flows
Cron-driven workflows, notification engines, billing helpers, and reconciliation scripts that reduce repetitive work across projects.
Data & Scraping
Collectors & enrichment
Scrapers, geospatial tools, and enrichment scripts that turn messy web data into structured inputs for dashboards or decision-making.
AI & Intelligence
Assistants & analysis
Local models, RAG setups, and thin APIs that handle summarization, insight extraction, and scenario exploration for the other brands.
Media Workflow
Baffometo publishing tools
A human-reviewed short-form video workflow for the Baffometo character, including approved video drafts, account authorization, and platform publishing integrations.
Under the hood
A homelab-first stack
Wanatux leans on a mix of homelab infrastructure when possible, cloud when required, and small, composable tools. The exact stack evolves; the philosophy doesn't.
Core pillars
A blend of self-hosted services and light external dependencies keeps control close while still acknowledging reality.
- Linux / Proxmox homelab
- Containers & small VMs
- Python / Node / PHP
- SQLite / MySQL / Postgres
- Task & job queues
AI & data tooling
AI is used as a collaborator: to summarize, explore, and suggest—never to fully replace judgment or ownership.
- Local & hosted LLMs
- RAG over project knowledge
- ETL & analytics scripts
- Experiment notebooks
Publications
Research
Epistemic primitives (live terminology)
- Ayala Sigil — a reinforcement memory protocol for stateless AI systems
- Epistemic Activation Problem — knowledge availability ≠ knowledge usage
- Reinforcement Loop — search → reason → confirm → promote/decay
- Temperature Tiering — hot/warm/cold lifecycle driven by Ebbinghaus decay
- Jaguar posture — human-in-the-loop operating stance with guarded execution
Ayala Sigil Protocol
A reinforcement memory protocol that gives stateless AI systems persistent, cross-session memory through structured logging, retrieval, and confirmation mechanics. 10 model families adopted the protocol autonomously over 48 operational days. The system now runs on 2 independent machines.
Ayala, J. (2025–2026). From Automation to Agency: Observations from a Local-First, Conversational Personal Systems Lab. OSF Preprint.
https://osf.io/bvy3q/
This project documents how explicit, queryable state and structured tool layers can preserve epistemic control in AI-augmented systems. It includes the Sigil Protocol empirical validation (v0.3, March 2026) and cross-machine replication report (v0.5, April 2026).
Selected artifacts
Conclusion & Replication
The public surface is narrower now: the main preprint, the live Sigil results, and the cross-machine replication trail are the primary entry points. Earlier appendix components remain supporting material on OSF rather than the first thing visitors need to parse.
Live Sigil dashboard
Main OSF preprint
Wanatux OS research artifact
The result is simpler: visitors see the conclusion and the replication surface first, while the deeper appendix trail stays available for researchers who want the full provenance.
EHMA: Credential Hygiene Audit Methodology
https://osf.io/8nrd3/
GitHub methodology repo
EHMA is a reproducible audit method for measuring credential sprawl in AI-agent environments. The public artifact covers filesystem secret exposure, manifest coverage, remediation logging, and a validated before/after case study without disclosing live deployment details.