The Sovereign Socratic: AI-Assisted Learning Is an Architectural Problem

A conversation with Mael Garbe—19-year-old Tech Architect, Founder of ORBITECH, and author of LBO Journal—on Socratic LLM engineering, the dual-engine strategy that pairs a commercial AI education ecosystem with a non-profit mission of free access to knowledge, and the operational convergence of software architecture and corporate finance.

By Louise Servoin · 2026-06-04 · 10 min read

## Mael Garbe – Founder of ORBITECH and Author of LBO Journal

Founder of ORBITECH and author of LBO Journal, a personal editorial project. Areas of expertise: AI architecture, sovereign cloud, private equity, and full-stack development. Based in Lille, France.

Some young technology builders understand how to write clean, production-ready code. Others dedicate their time to analyzing the complex financial engineering of Leveraged Buyouts (LBOs) and small-business acquisitions. It is exceptionally rare to find a nineteen-year-old operating fluently at the exact intersection of both.

Based in the Hauts-de-France region, Mael Garbe is establishing a unique profile as a "Searcher-Developer". As the sole developer behind ORBITECH’s two distinct entities—a commercial micro-enterprise building its AI education ecosystem and a non-profit association devoted to free access to knowledge—and the author of LBO Journal, Mael Garbe is reframing how we approach AI-assisted training, digital sovereignty, and traditional small-business modernization.

We sat down with Mael to discuss the dangers of "convenience-first" educational AI, the strategic separation of a commercial micro-enterprise and a non-profit association, the realities of hosting sovereign cloud infrastructure, and how writing code mirrors the structural math of a corporate buyout.

## The Socratic Engine

Mael, you originally launched your company as a B2B automation agency in early 2025, but quickly pivoted toward education in April of that year. What structural friction did you observe in the market that prompted this shift?

The pivot was a direct response to how students were interacting with mainstream generative AI. When conversational assistants first gained mass adoption, students immediately began using them to bypass critical thinking. They copied a homework question, pasted it into an LLM, and copied the direct answer back onto their sheets.

Technology is not valuable if it encourages cognitive passivity. If you build systems that reward easy convenience, users will stay dependent.

To fight this, we structured our core engine around the Socratic method. The platform's AI tools are designed to serve as decision accelerators, not final decision-makers. The system is built with strict pedagogical guardrails that prevent it from giving direct answers. Instead, it asks targeted, step-by-step questions to guide the student's reasoning. The learner must remain the pilot of their own thoughts.

My background as a youth counselor (Animateur BAFA) at Pévèle Carembault taught me a fundamental truth: real pedagogy is active. You don't teach by lecturing; you teach by engaging.

"Technology is not valuable if it encourages cognitive passivity. If you build systems that reward easy convenience, users will stay dependent." — Mael Garbe

## The Dual-Engine Strategy: Two Legal Entities, Two Missions

You operate two distinct legal entities under the ORBITECH name—a commercial micro-enterprise and a non-profit association. How do these two structures divide the work?

ORBITECH brings together two distinct legal entities with two complementary missions—deliberately kept separate, both legally and operationally.

The micro-enterprise is the commercial entity. It develops the AI ecosystem and the Socratic method for the French education market: the tools, the platform, the school-facing offer. That is the business side of the project.

The non-profit—a loi 1901 association, currently being deployed—carries a different mission: free access to knowledge, with no financial barrier, for the widest possible audience. Where the micro-enterprise serves the market, the association exists to keep learning open to everyone.

What unites the two is a single pedagogical conviction: the learner must stay active. Whether someone arrives through the commercial offer or the free association, the goal is the same—guiding reasoning step by step rather than handing over ready-made answers.

## Sovereignty Over Convenience

You build and maintain your entire technical stack as a solo developer. Why did you opt for a sovereign European infrastructure over more convenient, global server setups?

It goes back to the core principles of our manifesto: sovereignty over comfort. When you handle educational data, especially for minors, compliance is not a marketing checkbox; it is a structural necessity.

Concretely, 100% of our user data and backend run in France—on AWS in Paris (eu-west-3), via Supabase—under a strict GDPR framework, with Data Processing Agreements (DPAs) and Standard Contractual Clauses (SCCs), and with no user data ever used to train models. For inference, we use the Anthropic API today: it is the most technically viable option at this stage. The important point is that this is American compute power, not a transfer of data or of sovereignty. In parallel, a French inference solution is currently being evaluated.

It is far easier to route API calls globally without thinking about where the data lands, but building a local, GDPR-compliant cloud is about taking long-term responsibility. This architectural integrity is what allowed us to be covered by regional media—Ici Nord (formerly France Bleu Nord) featured the project in March 2026 under the title "Quand l'IA fait réfléchir les élèves" ("When AI Makes Students Think"), with France 3 Nord also reporting on the project, and to begin piloting the platform in private secondary schools in northern France.

## The "Searcher-Developer" Paradigm

You write LBO Journal, breaking down the financial mechanisms of small-business buyouts and due diligence. How do you connect the logical world of database engineering with corporate acquisitions?

They are deeply synergistic, which is why I view my path through the lens of a "Searcher-Developer". LBO Journal began as a regular editorial project—buyout analyses and due-diligence breakdowns, at a sustained pace during its launch phase. As ORBITECH grew—the Socratic method, the AI ecosystem, the launch of the association—I naturally eased off that cadence. Today I publish when the subject and the time align. It is a personal passion, distinct from ORBITECH’s core purpose, but one that directly feeds the Searcher-Developer posture.

Many traditional B2B service companies in regional France are managed by owners approaching retirement, but these business operations are often weighed down by severe technical debt. They rely on outdated, manual IT workflows, fragmented Excel files, and paper-based tracking.

If you approach these business acquisitions solely as a financial investor, you look at leverage ratios—essentially capping senior debt relative to EBITDA—and you calculate conservative Debt Service Coverage Ratios to ensure the cash flows can cover the acquisition liabilities.

But if you also have an engineering background, you can audit the company's workflows as a system of data pipelines. The moment you acquire the SME, you can step in and automate the administrative backlog, rewrite their database relational mapping, and implement lightweight, custom tools. By modernizing the software layer, you instantly optimize their operating margin, accelerating the rate at which the business pays down its debt. Writing code and structuring a leveraged buyout are both exercises in designing self-sustaining, optimized machines.

## The Question of the Future

As AI tools continue to commoditize standard code and writing, what is the one human capacity that an algorithm will never be able to automate?

The algorithm can calculate financial ratios or write PostgreSQL schemas, but it cannot navigate human risk tolerance or empathy-driven negotiations.

It cannot sit in a room with a retiring business owner who has spent thirty years building their local SME, understand their personal anxieties about the transition of their legacy, and build the deep personal trust required to close a deal.

At ORBITECH, our AI is a co-pilot designed to reduce cognitive load and challenge student reasoning, but the destination must always remain human-led. The ultimate measure of your career will never be an automated promise. It will always be the physical, operational trace of what you had the courage and discipline to actually build.

## The Golden Thread

We often see a gap between what technology can do and what professionals actually adopt. How do we ensure AI becomes a 'co-pilot' that elevates human judgment rather than a tool that encourages intellectual passivity?

For me, the answer comes down to a single design conviction: the Socratic method. I believe it is the central axis of the future — and in high-stakes fields like education and medicine, it may be the only genuinely viable one.

The logic is simple. A system that hands you the answer trains you to stop thinking. A system built on Socratic questioning does the opposite: it forces reflection. You don't receive a conclusion, you reach it. Through a structured exchange of questions, the user advances step by step, tests their own reasoning, and arrives at a response they actually own. That is the difference between consuming an answer and understanding a problem — and understanding is the only thing that carries over to the next problem you face.

This is precisely where AI earns the title of co-pilot rather than replacement. Working through a dense body of material on your own takes time; a Socratic AI compresses that curve. It accelerates the path to comprehension without short-circuiting it. The speed gain is real, but it serves integration, not avoidance — the learner still does the cognitive work, simply faster and with better guidance. Velocity in the service of autonomy, never as a substitute for it.

So the golden thread is this: build the system so that the human always remains the one who judges. The AI challenges, prompts, and accelerates — but the destination stays human-led. The moment the machine begins deciding for you instead of thinking with you, you have lost the very thing that made the tool worth building.

Tags: AI in Education, Socratic Learning, EdTech, Sovereign Cloud, Private Equity, LBO, M&A, Lille, Entrepreneurship