01
Define
Translate the problem into explicit requirements.
NDZ LABS · INDEPENDENT R&D STUDIO
Designing AI systems that make humans better at complex work
NDZ Labs is an independent AI R&D lab exploring how agentic systems can augment human work from product development and research to analysis and decision-making.
From product definition to implementation, testing, recovery and release — with humans making the decisions that matter.
02 — The Core Idea
AI doesn't have to make humans unnecessary to make them more productive.
We believe the most valuable AI systems are not necessarily the ones that replace the most human work, but the ones that remove the mechanical parts of complex workflows and give people more room for creativity, judgment, strategy and decision-making.
By delegating execution to agentic systems, humans can spend more of their limited attention on the parts of the process where their involvement matters most.
03 — The Question
AI can already execute many parts of complex workflows.
The harder question is what we should ask it to do
- How much autonomy should an agent have?
- When is a specialized team better than a single agent?
- When does automation actually improve human work?
- And what happens when execution becomes abundant?
NDZ Labs explores those questions by building real systems and testing them in real workflows.
04 — Systems we've built
Different questions require different architectures.
The NDZ Development System is an agentic development environment originally built to accelerate internal product development.
It evolved into a general-purpose system for coordinating product definition, design, implementation, testing, debugging and documentation across an entire software lifecycle.
02
Contract
Create a structured interface between roles.
03
Build
Allow an agent to execute within those boundaries.
04
Verify
Evaluate the result against explicit criteria.
05
Recover
Handle failure without restarting the entire process.
06
Evolve
Feed what was learned back into the system.
07
Produce a usable outcome.
A functional MVP emerges together with the documentation and history required to understand how it was built and how it evolved.
05 — The Agentic Architecture
There is no single architecture for agentic systems.
The right structure depends on the work being delegated, the level of autonomy required and where human judgment needs to remain in the loop.
NDZ Labs experiments with different architectural patterns to understand what works, where they fail and when one approach is preferable to another.
Declarative systems
Explicit responsibilities, contracts and workflows define how agents interact and what each stage is expected to produce.
Skills-based agents
A capable agent can acquire specialized behaviors through skills without requiring an independent agent for every responsibility.
Agentic teams
Multiple specialized agents can reproduce organizational roles such as Product Management, Design, Development and QA.
Human-led systems
The human remains an explicit component of the architecture, defining direction, making critical decisions and intervening where autonomy stops being useful.
The architecture is part of the experiment
06 — The Agnostic Layer
The intelligence can change. The system remains.
NDZ Labs systems are designed to remain independent from the environment in which they operate.
Models, providers, development environments, tools and infrastructure can change without redefining the system itself.
This separation allows us to move experiments across different environments without rebuilding the architecture around a specific vendor or workflow.
More importantly, it keeps the research focused on the system and the problem it is designed to solve, rather than on the particular technology used to run it.
LLM agnostic
The development workflow is not built around a specific model provider.
Different models can execute different parts of the system without redefining the product process.
IDE agnostic
The system does not depend on a particular development environment.
The orchestration layer lives above the tools used to write the code.
Language agnostic
The contracts describe what the system needs to achieve, rather than prescribing a specific programming language.
The implementation stack can change without changing the underlying development methodology.
Product intent
Contracts
Agent orchestration
Any model · Any IDE · Any language
Validated software
The implementation is replaceable. The system of coordination is not.
07 — Experiments
Every NDZ Labs project starts with a question
AerOS Personal
The first complete product built through the NDZ approach.
A portable, declarative contract system for governing product development.
Explore the product →Webeloper
Can an AI system allow someone without technical skills to run the complete sales and delivery workflow of a website business?
Explore the project →03 — More experiments in development
New systems are built not only to ship software, but to continuously test and improve the development infrastructure itself.
08 — The R&D Loop
Build. Measure. Learn. Rebuild the system.
Every product becomes an experiment.
Every experiment improves the system.
Every improvement changes what can be built next.
Build product
Observe the process
Identify bottlenecks
Improve the system
Build faster / better
Next product
NDZ Labs is not trying to predict what the future of software development will look like.
It is building it one product at a time.
09 — About / Founder
Built by Francisco Fernández
Francisco is an AI Product Architect and product builder with 6+ years of experience taking digital products from definition to production.
His work sits at the intersection of product strategy, software architecture, AI and hands-on development.
NDZ Labs started as a personal R&D environment to explore faster ways of building products. The development system emerged from that work — first as an internal tool, and later as a broader architecture with potential applications beyond NDZ itself.
Have a product worth building?
Bring us the problem.
We'll help define the system, find the fastest path forward and build what comes next.
Start a conversation →NDZ Labs
Independent AI Product R&D
Buenos Aires · Working globally