PEOPLE + AI · SYSTEMS WITH MEANING

Build with AI.
Keep the thread.

From the first idea to an evolving system. We are building tools that keep intent, meaning, and evidence connected — with decisions in the hands of the people responsible for the work.

Explore the products From domain to code
The proposition, in a loop
01

Intent

What do we want to achieve?

02

Meaning

What are we talking about?

04

Evidence

Does it meet expectations?

03

System

What did we build?

Observe. Compare. Learn.

The next step remains a decision.

Shared meaningTraceable evidenceHuman control

01 / THE THESIS

The work continues.
The context should too.

A conversation holds the idea. A document holds the decision. The code holds the implementation. Between each stage, someone has to reconstruct why it all exists.

Our thesis is that people and machines can build better when work preserves these connections.

02 / THE PRODUCTS

Different pieces.
A shared intention.

Construction, language, reasoning, and a place to work. Each product addresses a different part of lost context.

CraftAlchLangWorkspaceOntology ProductEssaysAgent Reasoning Engine

Craft

System construction

In development

From what a product needs to a system we can verify.

Craft connects product intent, domain meaning, and construction. Its proposition is to make implementation choices explicit and compare observed behavior with expectations.

What we want to make possible

Start with a feature, define its concepts and rules, build a realization, and examine the evidence before continuing or correcting it.

What we aim to preserve

  • Explicit architecture decisions
  • Implementation connected to the domain
  • Evolution informed by evidence

AlchLang

Domain language

In development

A shared language for understanding and building.

AlchLang expresses domain concepts, relationships, rules, states, and events. It gives people and tools an explicit representation of the model to inspect and discuss.

What we want to make possible

Describe what makes a unit sellable and which events change its state, while keeping the vocabulary recognized by domain experts.

What we aim to preserve

  • Explicit concepts and relationships
  • Rules open to inspection
  • Meaning independent of a tool

Workspace

A place to work

In development

A conversation that stays with the work.

Workspace proposes a continuous space for conversation, artifacts, and delegated work. Attention can move from text to a graph without asking the person to reconstruct the context.

What we want to make possible

Request research, follow the work, open the resulting text, and inspect its sources within the same experience.

What we aim to preserve

  • Continuity across tasks
  • Attention on the active artifact
  • Intervention when a decision is needed

Ontology Product

Connected meaning

In development

Meaning belongs to the project.

The Ontology Product proposes organizing concepts and relationships in models connected to their sources. Meaning should be queryable, testable, and able to evolve without handing ownership to a tool.

What we want to make possible

Trace a rule back to its defining source, inspect its relationships, and make contradictions or open questions visible.

What we aim to preserve

  • Connections to sources
  • Concept identity
  • Inspectable changes and uncertainty

Essays

Writing with continuity

In development

The text evolves. The intent stays.

Essays proposes a focused environment for substantial writing, with sources, revisions, and editorial decisions connected. The tool supports writing while the author keeps judgment.

What we want to make possible

Develop an essay, receive feedback on specific passages, choose which revisions to accept, and export clean text for its audience.

What we aim to preserve

  • Author intent
  • Findable sources and feedback
  • Preserved editorial decisions

Agent Reasoning Engine

Reasoning and context

In development

Continue from what was understood.

ARE aims to preserve the objective, interpretation, evidence, and unresolved work across interactions. It compares expectations with outcomes and investigates improvements through research and controlled experiments.

What we want to make possible

Resume a task knowing what was tried, which evidence changed the approach, and what remains to be resolved.

What we aim to preserve

  • An explicit current objective
  • Context with identifiable sources
  • Learning limited to its scope

03 / HOW WE THINK

The outcome matters.
So does the path.

A work loop connecting intent to action and action to evidence. Select a stage to explore the proposition.

01Understand02Build03Verify04Evolve

Begin with meaning.

Make the desired outcome, domain concepts, and unanswered questions explicit.

An interpretation to discuss before a solution.

Make choices visible.

Connect intent to construction decisions, with a clear scope for the work of people and agents.

A realization we can explain.

Compare with expectations.

Observe the outcome and examine the evidence. A test can support an expectation, contradict it, or leave a question open.

Evidence for choosing the next step.

Continue without starting over.

Preserve decisions, lessons, and responsibility for the next step. Adjust when evidence calls for it.

Continuity with room for correction.

WHAT GUIDES OUR CHOICES

People keep control.

Automation should make decisions understandable and correctable.

Sources stay close.

A claim should remain connected to what supports it.

The gaps stay visible.

A polished interface should not conceal open questions.

04 / AN IDEA IN PRACTICE

“In stock”
for whom?

A returned unit. A pending inspection. A reservation confirmed too soon. Follow a DDD story and discover how distinguishing words changes the code.

Explore the presentation ILLUSTRATIVE CASE · INTERACTIVE PRESENTATION

05 / THE LAB

Investigate.
Build. Question.

Vladimir RondelliResearcher and architect in AI governance

CyberAlchemy Lab brings together research and product development for work between people and AI. Theses guide what we pursue; prototypes, use, and evidence must show what works.

The propositions presented here are in development. The lab explores how to preserve meaning, responsibility, and human control throughout the work.

Questions guiding the research
  • How can we reduce context reconstruction without hiding important decisions?
  • What evidence shows that a tool improves the work?
  • How can models and tools remain replaceable without losing meaning?