Is Swoft Deep Tech? Strict application of Hello Tomorrow's five criteria

Many companies declare themselves Deep Tech. Few accept the test. This article applies, with all due rigour, the five Hello Tomorrow criteria — engineering challenge, science fundament, capital intensive, long timelines, transformative impact — to Swoft. The goal isn't to claim a label but to honestly check which one we deserve.

Criterion 1: engineering challenge

The product cannot be built by assembling available bricks. You must solve a complex engineering problem nobody has yet solved.

Verdict: yes. Building a platform that combines a business-domain metamodel (DDD), a substantial Event Store, AI agents architecturally bounded by bounded contexts, Rust code generation driven by that metamodel, a quality layer (CQL) that rejects non-conforming code before compilation, and all of this with measurable systemic coherence, is not a trivial assembly. No existing framework combines these seven elements. LangChain orchestrates, but without bounded context. CrewAI organizes, but without Event Sourcing. Datadog measures, but does not generate. Stacking these components took two and a half years of continuous R&D, which validates the engineering challenge.

Criterion 2: science fundament

The technology rests on cutting-edge scientific research — peer-reviewed papers, foundational patents, doctoral theses.

Verdict: partially. Swoft does not invent the theories it builds on. Domain-Driven Design comes from Eric Evans (2003), Event Sourcing from Martin Fowler and Greg Young (2005-2010), CQRS from Bertrand Meyer (1988), multi-agent systems from Ferber (1995), neurosymbolism from Henry Kautz (AAAI 2022). What Swoft contributes is the integration of these theories into an operational system — itself an open research question.

The 2025 systematic review of neurosymbolism by Colelough and Regli, covering 167 papers, identifies four unresolved gaps in production neurosymbolic systems: replay determinism, generality of the symbolic component, scalability of reasoning, multi-agent accountability. Swoft addresses all four — the AI decision stored as immutable event solves determinism, the generic metamodel solves generality, localization by bounded context solves scalability, dual attribution solves accountability. That is a scientific contribution in the strong sense.

The criterion is met, but you have to be honest about the level. Swoft is Applied Deep Tech — we industrialize existing research results into a production-ready system. It is not Foundational Deep Tech in the Pasqal sense, which solves open problems in theoretical physics. This distinction matters.

Criterion 3: capital intensive

Development requires substantial investment, often tens of millions of euros over several years, before first revenue.

Verdict: partially. On this criterion, Swoft sits between software Deep Tech and classic SaaS. The capital required to develop the platform — two and a half years of continuous R&D by a team of senior engineers and AI agents, plus cloud infrastructure, plus LLM model fees — runs to millions of euros, not tens of millions. It is more capital-intensive than classic SaaS but less than hardware Deep Tech (gigafactory, semiconductor fab, quantum computer).

The criterion is therefore met under a broad definition. It is not met in the hardware-Deep-Tech sense. This is one of the points where Swoft deserves the label software Deep Tech rather than Deep Tech tout court.

Criterion 4: long timelines

Technological maturation takes time, five to fifteen years between initial R&D and commercial scale.

Verdict: yes. Swoft R&D started in 2023. The product is production-deployable in 2026 in regulated sectors. Its commercial maturity at scale will not arrive before 2027-2028, the time to structure the customer base, stabilize integration standards and absorb the learnings. This 5-6 year timeline between initial R&D and commercial scale is consistent with the Hello Tomorrow criterion.

Criterion 5: transformative impact

If the company succeeds, the impact far exceeds the business case: it changes an entire sector, or even opens a new category.

Verdict: yes, if the promise is delivered. Swoft promises to replace the standard SaaS model for critical domains with a model of bespoke software generated under architectural constraints. If that promise holds, it is a category change: you no longer choose between a rigid SaaS that doesn't quite match the business and long, expensive custom development. You get both at once — bespoke and industrial.

The transformative impact extends across several dimensions. For client companies, it is the end of the trade-off between standardization and adaptation. For the software market, it is the emergence of a third category between SaaS and custom development. For European digital sovereignty, it is a platform that industrializes agentic AI without dependency on a single vendor. For regulators, it is a natively auditable system. The promise, if delivered, does transform the sector.

Difference between Foundational and Applied Deep Tech

This distinction is not a deflation of the claim. It is a useful category that many companies confuse or avoid. Foundational Deep Tech invents new physics, new chemistry, new biology. Applied Deep Tech turns known scientific results into industry. Both are valuable, but they don't have the same risk, capital and timeline profile.

Pasqal is Foundational Deep Tech: it solves open problems in neutral-atom physics. Mistral is Applied Deep Tech: transformers have existed since 2017, but industrializing them into an LLM competitive with GPT and Claude is a major engineering and industrial challenge. Neither is less legitimate than the other — they are simply different.

Swoft clearly fits the second category. DDD has existed for 20 years, Event Sourcing for 15, neurosymbolism for 30. What we do is combine them into a production-ready system, an open question nobody had answered. That is Deep Tech in the strong sense, but applied.

Why this precision matters

Three reasons make this precision useful, for us as much as for our interlocutors.

First, it avoids inflating the promise. A company that declares itself Deep Tech without applying the criteria exposes itself to a reality check from specialized investors and public decision-makers. Rigour protects credibility.

Second, it clarifies what we offer. Swoft is not a scientific breakthrough. It is a platform that industrializes rare techniques in production. That promise is different from that of a Foundational Deep Tech — more pragmatic and more immediate. For a client looking for a reliable enterprise system that holds over time, it is more useful than the promise of a future revolution.

Third, it places Swoft in the right investment box. European software-Deep-Tech funds have specific criteria — less capital-intensive, shorter timelines, faster exits. That is our natural category, not hardware or Foundational Deep Tech.

We do not build a new science. We turn known sciences into industry. It is less glamorous than fusion or quantum, but it is what changes the life of a bank, a medical practice or a hotel tomorrow morning.