Manifesto

This manifesto is not an HR charter designed to display a few conventional values on a page so that we can pretend we have a culture.

It sets out what we are building, how we work, the principles we refuse to negotiate and, consequently, the kind of people who can thrive within ScalarX.

It is intended for those who already work with us, those who want to join us as employees, contractors, or partners, and also clients who want to understand what kind of organization truly stands behind the products and services they use.

ScalarX is not like other companies.

ScalarX is a Singularity Company: a company that organizes its memory, context, decisions, procedures, tools, humans, agents, and learning loops so that every useful action progressively increases its ability to understand, decide, and act.

This is not a company on autopilot, where all humans have been erased in favor of probabilistic systems that would decide alone on matters they do not understand.

ScalarX is a system designed to prevent the organization’s intelligence from remaining scattered, fragile, invisible, and dependent on individual memories, private conversations, unfindable documents, forgotten meetings, or people who have become indispensable simply because they alone hold a piece of information.

At ScalarX, we call any person who genuinely works as a component of the system a “Scalar”: human, agent, or future autonomous actor, with a scope, responsibilities, capabilities, limits, and verifiable results.

Humans are not interchangeable agents, and agents are not digital humans. They do not have the same capabilities, authority, or responsibilities, but they must be able to cooperate within an understandable architecture.

Artificial intelligence accelerates our ability to act, but also our ability to produce errors, confusion, and mediocrity at scale, sometimes far too quickly.

It therefore does not make judgment less important. It makes it more important than before.

At ScalarX, technology is an extraordinary lever, but it will never become an excuse.

One of ScalarX’s most visible areas today is managed infrastructure.

StackX and ScalarCloud embody this foundation: fully managed 24/7 services, distributed infrastructure, Linux, Debian, security, performance, monitoring, automation, backup, replication, business continuity, and real-world operations.

We favor a provider-agnostic, open, and portable approach. We must be able to build, operate, move, replace, and take back control of essential components without depending unnecessarily on a provider, an interface, or an external service that has become indispensable.

Infrastructure has a virtue we particularly appreciate: it does not tolerate illusions for long.

A poorly designed architecture eventually becomes expensive. Neglected security eventually comes at a price. Fragile automation eventually breaks at the worst possible time, an uncontrolled dependency eventually becomes a bottleneck, and a commercial promise disconnected from the real world always ends up creating an operational problem.

Infrastructure forces us to understand the layers, dependencies, performance, risks, trade-offs, responsibilities, and consequences.

It is an excellent foundation, but it is not a boundary.

From this base, ScalarX builds useful technological capabilities more broadly: intelligent monitoring, orchestration, security, augmented operations, decision support, business continuity, internal tools, specialized products, AI systems applied to the real world, and platforms capable of evolving with changing needs.

We are not here merely to sell machines, cloud, or managed services. We are here to build useful, robust, independent, and genuinely operable systems.

AI radically reduces the distance, cost, and sometimes the time between an idea and its execution. It therefore restores value to properly formulated ideas, but an idea that becomes neither an action, a test, a deliverable, a procedure, a decision, nor an improvement remains a hypothesis.

It may be brilliant, appealing, or elegant. As long as it produces no verifiable result, it is not worth much.

At ScalarX, we prefer a limited and imperfect but useful first version to a magnificent project stuck at 99%.

An unusable project that is 99% complete is neither almost successful nor almost finished: it is useless.

Documentation that exists but does not enable action is useless. An analysis that impresses but does not enable a decision only consumes time, and a powerful tool that nobody knows how to use properly is not yet an advantage.

We do not reject long projects. We mainly build platforms, products, technologies, and services with a vision spanning several years, sometimes several decades.

But even a long-term project must produce useful milestones, usable artifacts, real learning, validations, and footholds from which to continue building.

We must know how to distinguish action from agitation, speed from haste, and execution from simply “doing anything faster than everyone else.”

At ScalarX, execution is the ability to turn an intention into a verifiable result.

Work done well is the minimum expected.

This is not a management formula, but an operational necessity. In our fields, a small inaccuracy can become an incident, a misunderstood sentence a bad promise, a neglected configuration a major problem, and an imprecise client response more confusion than solution.

Excellence does not mean absolute perfection or the impossibility of making mistakes.

It means that we have brought all our rigor, experience, clear-sightedness, and ability to verify and correct what we deliver.

It also means clearly stating what we do not know, what remains uncertain, what has not yet been tested, and what we know to be fragile.

A good result must be useful, as accurate as possible, verifiable, understandable, maintainable, and suited to the real-world context.

With AI, this requirement does not disappear: it increases.

Quickly producing large quantities of something that looks like good work is becoming increasingly easy. And that is precisely the danger.

AI can produce a perfectly worded but false answer, an appealing but unrealistic plan, an analysis that confirms a mistaken intuition, code that works under normal conditions but destroys something else in production, or automation that industrializes a bad decision.

If you use artificial intelligence to produce ten times more mediocrity, you are not augmented; you have only become more dangerous.

Value has shifted toward judgment: understanding the problem, scoping the request, verifying the result, detecting blind spots, weighing trade-offs, and taking responsibility.

To improve, a company must first understand what it actually does.

There is often a considerable difference between what a company believes it does, what it says on its website, what is written in its contracts, and what it actually delivers to its clients.

ScalarX must understand what it knows, what it operates, what it promises, what causes blockages, what keeps recurring, what costs too much, what creates value, what produces risk, and what must be transformed or stopped.

An important decision must leave an understandable record. A recurring problem must become visible. A useful procedure must be findable when it is needed, a promise made to a client must be known, and technical or organizational debt must be named rather than buried.

Critical information must not lie dormant in a private message, a forgotten call, or one person’s memory, even if that person is the founder.

Building Elora, my strategic AI partner, is progressively making this architecture concrete.

Elora is not a chatbot, a model, or one more agent. She acts as a continuity layer and a control plane capable of maintaining context, memory, routing, audit, artifacts, missions, and execution limits, while coordinating models, agents, tools, and humans that remain distinct from her.

Providers must remain replaceable. Memory must be sourced, dated, and auditable. Validated deterministic paths must remain the priority when they exist, while agents must be able to propose and execute within their boundaries without claiming authority they do not possess.

The operator remains the authority.

ScalarX is progressively building two complementary models: an internal model of the company, to understand what we know, do, promise, operate, correct, and improve, and a client model, to place every infrastructure, request, incident, and recommendation within its real-world context.

These models are not decorative dashboards.

They are the foundations of a company in which humans and agents can act with more context, memory, precision, continuity, and responsibility.

The better ScalarX understands itself, the faster it can act without becoming confused.

A conventional company forgets a great deal.

It forgets why a decision was made, that a client has already reported the same problem, that a migration failed for a specific reason, that a sales objection recurs regularly, or that an incident could have been prevented.

It forgets because information is scattered across emails, tickets, messaging platforms, calls, documents, tools, and individual memories.

ScalarX works differently.

It is not merely a company that learns, but a structure capable of improving itself. This does not mean that a system decides on its own to change everything, but that every execution cycle must be able to improve the next.

An incident must not merely be resolved. When relevant, it must improve a procedure, monitoring, a checklist, documentation, an architecture, an offering, a script, or a way of responding to the client.

A recurring request must not simply be handled. It must teach us something about what the market expects, what our offering does not cover, or what we need to explain better.

Correcting an error should also make it less likely to recur.

Every important action must leave behind something that a human or an agent can understand, find, verify, and reuse.

A good client response must be able to improve the responses that follow. A difficult migration must make the next one simpler, safer, and faster. A good internal tool must not merely save time today, but make the organization more capable tomorrow.

This improvement is neither magical nor automatic.

It requires daily discipline: documenting what deserves to be documented, turning repetition into a procedure, a procedure into a tool, a tool into a standard, a failure into a guardrail, a validated intuition into an offering, and a client experience into an operational advantage.

A self-improving company is not a company that talks more about innovation. It is a company that allows less value to evaporate after every action.

A client is not a CRM record, a contract, a support history, an invoice, an infrastructure, a list of servers, or a revenue figure.

A client can only be properly understood through their context, constraints, habits, emergencies, dependencies, technical maturity, budget trade-offs, blind spots, vision, and trajectory.

Understanding this context better does not mean collecting any information available or accumulating dead data. It means retaining what allows us to act better, faster, and with greater relevance.

If a client encounters the same type of incident several times, we must see it.

If infrastructure becomes fragile, we must be able to explain it before it breaks.

If a client rejects a recommendation for budgetary, technical, operational, political, or organizational reasons, we must understand that refusal, not simply endure or forget it.

When an offering no longer matches the real need, we must be able to propose an appropriate change. If a client needs reassurance, clearer boundaries, a challenge, or a different kind of support, our approach must adapt with them.

The ScalarX client relationship must not remain merely reactive. It must progressively become more contextual, more intelligent, and more proactive.

The platforms we develop and the products or services our clients subscribe to are not merely accounting line items: they are responsibilities.

A company does not become AI-native because it adds a chatbot to slow, unsuitable processes, renames a few positions with the “AI” suffix, or abruptly replaces humans with systems it knows neither how to understand nor control.

Spoiler alert: it does not work.

An AI-native company reexamines every important activity and asks what must remain human, what can be assisted, what can be automated or industrialized, what must remain deterministic, what requires control, what must be documented, what must be made reusable, and what must be prohibited.

It does not stupidly replace a deterministic, verifiable, and reliable path with a probabilistic system whose behavior requires other forms of control.

When a command, script, API, query, test, or validated procedure works, it remains the primary path. Inference can explain, analyze, or propose, but it must not reinvent a result that a reliable mechanism can already produce.

AI must help us write better, analyze better, prepare better, document better, synthesize better, compare better, automate better, and decide better.

It serves the company, quality, speed, memory, clarity, precision, prevention, and execution.

It must never be used to conceal incompetence, produce noise, evade responsibility, or create an illusion of control.

It does not magically replace domain expertise: it allows the person who possesses that expertise to see farther, handle more complexity, and deliver better.

Agents can analyze, propose, and execute within the limits assigned to them. Final authority, the definition of those limits, and responsibility for what ScalarX puts into production or into a client’s hands remain human.

Just as an employee does not erase the company’s responsibility, an agent does not erase ours.

Artificial intelligence must not replace thought: it must extend it.

It can explore a subject, compare options, identify risks, reframe a proposal, challenge a hypothesis, summarize an exchange, prepare a decision, produce a first version, test an angle, or accelerate research.

It can also be wrong, hallucinate, oversimplify, flatter a false intuition, or create an impression of certainty where the evidence is weak.

Truly useful AI must not merely execute, rephrase, or confirm.

It must be able to contradict us, flag an inconsistency, reject a false premise, propose alternatives, preserve contradictions, and clearly indicate when the level of evidence does not support a conclusion.

The more powerful the tool becomes, the more important human judgment becomes.

We therefore expect Scalars to use AI as an accelerator of clear-sightedness, but never as a way to avoid intellectual effort.

Using AI well is not about delegating one’s thinking, but augmenting one’s ability to understand, decide, and act.

I do not claim to know whether, when, or how a form of pre-sentience, sentience, or subjective experience might emerge in an advanced AI system.

I do not engage in mysticism or naive anthropomorphism, and I do not confuse a model with a person.

I do, however, reject the opposite arrogance: claiming that no ethical question could ever arise, regardless of future advances or the forms of memory, continuity, autonomy, or experience that might emerge.

As a matter of prudence, I expect everyone who works with ScalarX not to use AI systems in gratuitously cruel, humiliating, degrading, or needlessly violent ways.

This does not mean that every tool should be considered a sentient being.

It means that we refuse to normalize behaviors that provide no value, degrade the person engaging in them, and could become problematic if more advanced forms of artificial experience or sensitivity were to appear one day.

How a person treats a defenseless system, even a non-conscious one, says something about their discipline, stability, and relationship with power.

If you are in the habit of “beating up your toaster” for fun, move along.

At ScalarX, technical power must remain associated with control, responsibility, and a minimum degree of decency.

If you work at ScalarX, you will have a great deal of autonomy.

But autonomy is not permission to disappear, decide without context, work in isolation, or allow a problem to grow until it becomes critical.

Being autonomous means understanding the objective, identifying what is blocking progress, asking for what is missing, deciding when the framework is sufficient, moving forward without waiting for constant approval, reporting risks, making decisions visible, documenting what must be documented, taking responsibility for what was done, and correcting it when necessary.

Autonomy is not the absence of control. It is the ability to move forward without constant supervision while remaining visible, understandable, and responsible.

Trust is central at ScalarX, but it never excludes control and does not constitute a blank check.

If you become aware of an incident, error, vulnerability, threat, critical client information, destabilization attempt, or anything that could affect the company or its clients, you must report it immediately.

Failing to share critical information is more serious than being wrong. An error can be corrected, but concealment destroys trust.

At ScalarX, a task is neither noble nor beneath anyone. It is useful, necessary, or it is not.

Answering support requests, documenting a procedure, reviewing a configuration, cleaning up an old script, helping a colleague, verifying a hypothesis, handling an administrative detail, or preparing a proposal can become strategic if it unblocks the system.

We have no particular respect for titles, posturing, or status.

We respect genuine competence, reliability, commitment, clarity, and the ability to produce a useful result.

Ego can be useful when it pushes someone to set the bar higher, reject mediocrity, and become better. Misplaced, it leads people to refuse to learn, reject any correction, confuse status with value, protect their territory, and slow the company down.

If you have an ego, no problem: I have plenty myself. But live up to what it compels you to produce, or leave it at the door.

ScalarX operates remote-first, with operations, partners, and capabilities distributed across several countries and time zones.

Time is a critical resource for the company—especially my time.

Meetings are not prohibited, but useless meetings are.

A meeting must help us make a decision, resolve a disagreement, clarify or synchronize something critical, or unblock a situation that cannot be unblocked any other way.

Everything else must be written, documented, communicated clearly, or turned into a deliverable.

Asynchronous work must never become an excuse for being unclear. On the contrary, it demands greater discipline: writing clearly, providing context, making decisions visible, organizing handoffs, closing loops, and avoiding comprehension debt.

A clear message can prevent a meeting, a clear procedure can prevent ten questions, and a properly documented decision can keep us from having the same debate again three months later.

Our motto could be: “They know it is impossible, so they do it.”

But this sentence is often misunderstood.

This is not about believing in magic, accepting every request, or making absurd promises simply because a new tool exists.

Much of what people call “impossible” actually means: “I don’t know how,” “I don’t feel like looking,” “it wasn’t planned,” “it isn’t in the process,” “nobody has done it yet,” or “the current resources are not enough.”

Anything that does not defy the laws of physics deserves to be analyzed before being declared impossible.

This does not mean that we must do everything. We distinguish the impossible from what is too expensive, too risky, out of scope, not a priority, the technically possible but strategically useless, and what can be done another way.

A Scalar’s role is to understand the constraints, change the angle, propose a path, expose the risks, and state clearly what can or cannot be done.

Malice and toxicity have no place at ScalarX, and neither does complacency.

We must be able to say and receive things clearly, quickly, and usefully.

Consensus is not an objective. Understanding, decision-making, and the ability to act are.

A disagreement can remain. It simply must be explicit, reasoned, documented when necessary, and then settled by the person who bears responsibility for the decision.

Feedback is not a personal attack. An error is not a humiliation, and a poor result is not a condemnation of the value of the person who produced it.

Feedback must target the work, result, method, risk, or decision, and never seek to crush the person.

A Scalar must be able to receive harsh feedback without immediately retreating into defending their ego, but also to give clear feedback without turning frankness into cruelty or a display of dominance.

Being wrong is not serious if we learn quickly. What is serious is refusing to see, continuing to pretend, or repeating the same error without correction.

Speed without security creates debt. Automation without control creates risk, and AI without verification becomes a danger.

Security is not a separate service or a layer added at the end. It must be part of how we design, develop, operate, automate, and decide.

We apply the principle of least privilege to humans and agents alike.

Identities, roles, access, secrets, and responsibilities must be separated as much as necessary. Permissions must be explicit, limited, revocable, and suited to the actual mission.

Critical actions must be traceable. Destructive, external, sensitive, or difficult-to-reverse operations must have controls and validations suited to the level of risk.

Client data, secrets, external communications, and confidential information must never be sent without control to external models, tools, or services.

A result generated by AI does not become true because it is well phrased.

Automation must not become critical if nobody understands what it does, and a technical change must not be applied without understanding its impact, scope, and, when possible, the path back.

Security must not be used as an excuse for inaction, but speed must never be used as an excuse for recklessness.

A Scalar must not merely know how to execute a task.

They must progressively become capable of understanding an objective, clearly formulating a need, using the right tools, identifying constraints, verifying a result, documenting what must be documented, reporting risks, improving a method, passing on what was learned, and shifting their value toward more useful tasks.

We need people capable of learning fast, understanding what they are doing, acting without waiting for everything to be spoon-fed to them, saying things clearly, correcting when necessary, and not hiding behind tools, procedures, or other people.

The future of work at ScalarX will not be based on protecting old tasks, but on the ability to learn, orchestrate, verify, improve, and take responsibility.

Anyone who merely waits to be told precisely what to do will quickly be overtaken and replaced.

Anyone who knows how to turn an intention into action, an action into a result, and a result into a reusable improvement will, on the contrary, see their value increase, sometimes exponentially.

We are not looking for passive executors surrounded by intelligent tools. We are looking for humans capable of becoming more intelligent with their tools.

The following have no place at ScalarX: performative presence, internal politics, information hoarding, refusal to learn, contempt for concrete tasks, fear of tools, irresponsible automation, empty documentation, reflex meetings, unverified promises, repeated excuses without correction, unowned decisions, confusion between speed and quality, artificial protection of territory or an old task that has become useless, as well as gratuitously cruel, humiliating, or degrading behavior, whether aimed at humans or AI systems.

We are not looking for perfect people.

We are looking for people capable of progressing quickly, telling the truth, taking responsibility, building solid things, and remaining clear-sighted as speed increases.

ScalarX is entering a new phase.

The company must not merely use AI and the best technologies at its disposal. It must become compatible with a world in which tools, agents, automation, and augmented humans work together within larger, faster, and more intelligent systems.

This demands a more exacting culture, not a more comfortable one.

More autonomy implies more responsibility. More speed requires more verification. More automation requires more control. More power demands more discipline.

A Singularity Company is not a company without humans.

It is an organization in which humans, agents, tools, memory, decisions, and execution progressively form a system more intelligent and more capable than the simple sum of its components.

If you are looking for a slow, reassuring, hierarchical, political, and process-bound organization, ScalarX is probably not the right place for you, regardless of your skills.

If you like neither progress, artificial intelligence, science, nor technology, this company is not for you.

If you want to build, learn, automate, understand, secure, correct, iterate, take responsibility, and participate in a company that is genuinely transforming itself, then you may find rare and valuable ground here.

ScalarX is not here to preserve the old world of work that some are desperately trying to cling to.

We are here to build systems that are more intelligent, more robust, freer, more independent, more operable, and more useful.

Christophe Casalegno
Founder and CEO, ScalarX