Early-stage · Independent project · Argentina
NAOF — the Nahuel AI Orchestration Framework — is being developed to make AI-assisted workflows more traceable, controlled and auditable: clear rules before the work, recorded evidence during it, and a human decision at the end.
Independent founder · Santiago del Estero, Argentina · Self-funded, pre-commercial development
A clearly scoped request enters the workflow.
Limits and decision boundaries set in advance shape what can happen.
Outputs and checks are recorded so they can be inspected.
A person reviews the evidence and makes the decision.
AI models can draft code, summarize documents or propose answers in seconds. But a fast answer is not automatically a reliable one — and it is often hard to see how it was produced, what was checked, or who approved it.
NAOF is an engineering project being developed around explicit constraints and verifiable evidence, rather than treating model outputs as automatically trustworthy. Its first direction is accountable AI-assisted software engineering, with a long-term vision for enterprise and educational applications.
Think of it as a lab notebook for AI-assisted work: every step is bounded, written down, and signed off by a person.
NAOF's approach rests on a small set of principles that keep AI-assisted work bounded, inspectable and under human responsibility.
Structured workflow states, predefined decision boundaries and clearly scoped execution paths.
Records and checkpoints designed to make workflow decisions and outcomes inspectable.
Decisions that matter require explicit human approval. People keep the authority.
Automated processing is kept separate from the review of its results.
The control architecture NAOF is exploring, simplified into five stages.
Conceptual model. This shows how a controlled workflow is meant to behave — not a description of a deployed system.
Work begins as a clearly defined, scoped request.
Processing runs within limits and paths defined in advance.
Results are checked against explicit criteria.
Outputs and check results are captured for inspection.
A person reviews the evidence and approves — or doesn't.
Want to know which parts exist today? See Development status — the initial kernel and router are implemented; other capabilities are under development or planned.
The initial focus is software engineering. Other applications represent longer-term opportunities to explore — not products available today.
A proposed foundation for controlled AI-assisted development, code review, testing and technical documentation — with explicit checks, evidence and human approval boundaries.
Potential future directions
Potential orchestration of bounded business workflows, keeping actions observable and authorization separate from execution.
Possible workflows for structured investigation, document comparison, information extraction and traceable research outputs.
Exploring ways to verify outputs, record supporting evidence and keep consequential decisions under human responsibility.
A long-term aspiration to explore responsible AI support for teachers and students. Read the vision
These are development directions. NAOF does not currently offer commercial deployments or production-ready autonomous AI workflows.
NAOF is an early-stage engineering project, not a production-ready commercial service. The initial deterministic, local read-only kernel and router have been implemented. Further shadow-workflow capabilities are under incremental development and evaluation.
NAOF is not yet a commercial service. Product-market validation, AI-provider integrations and production deployment remain future work.
NAOF plans to evaluate Claude as one potential model provider for bounded AI-assisted software engineering workflows.
A proposed proof of concept would receive a structured task, invoke an AI model under defined limits, capture outputs, apply validation checks and prepare evidence for human review. This integration is planned research, not a feature currently implemented or deployed.
Provider-extensible by design. Claude is one candidate to evaluate. NAOF is designed to remain extensible across model providers.
Beyond enterprise applications, NAOF's founder hopes to explore responsible AI tools that could help teachers and students in Santiago del Estero, Argentina.
AI should support teachers, not replace them.
Possible future uses
There are no school deployments or institutional partnerships at this stage. Any future educational pilot would require institutional involvement, appropriate safeguards for minors, student privacy, accessibility and curricular alignment.
Human teachers remain responsible for educational decisions.
Begin with small, evaluated pilots only when the necessary safeguards and partners are in place.
Investigate how AI could improve access to individualized learning assistance in Santiago del Estero.
The aim is to turn a rigorously developed engineering framework into useful technology for teams and organizations: validate software engineering workflows first, then explore broader enterprise applications where traceability, accountability and human control matter.
Alongside a sustainable technology business, the founder hopes to contribute to educational innovation in Santiago del Estero. These remain long-term goals, not current commercial offerings or educational deployments.
Verifiable outcomes over opaque claims.
Bounded scope, tests, documented limitations and iterative evaluation.
Technical validation, product discovery and future collaboration opportunities.
NAOF is an independent, pre-commercial AI orchestration project. Conversations with researchers, engineering teams, educators and prospective collaborators are welcome.
Use the email address with any mail app you prefer, or one of the shortcuts.