Deep analyses for CTOs and engineers: what is a real AI agent, how to think about multi-agent systems, what is Deep Tech, how Swoft sits in neurosymbolic state of the art.
ChatGPT isn't an agent in the academic sense. Russell, Norvig, Wooldridge and Ferber have proposed for 30 years a demanding definition that most 2026 products don't meet. State of the question.
ChatGPT wrappers, no-code platforms, code assistants, orchestration frameworks: everything labels itself as an AI agent. Confronting the academic definitions with real market products.
The Swoft definition strictly applies Wooldridge's four properties and adds three architectural constraints that make the agent reliable in production: bounded scope, native traceability, organizational identity.
Jacques Ferber wrote Multi-Agent Systems in 1995. Thirty years later, his organizational dimension — often forgotten by contemporary frameworks — becomes production-ready thanks to event-driven architecture.
The term Deep Tech has covered everything and its opposite since 2014 — quantum, biotech, fusion, AI, blockchain. Let's go back to the canonical definitions of BCG, Hello Tomorrow and Bpifrance to distinguish what is from what isn't.
Mistral, Pasqal, Aledia, Verkor, Diabeloop: France has more Deep Tech than most European countries but struggles to turn these gems into global champions. State of the ecosystem in 2026.
Rather than claiming the label, we apply it. Five criteria, five honest tests. Verdict: Applied Deep Tech, not Foundational Deep Tech — and we explain the difference.
Lovable, Bolt, v0 on one side. Architected platforms on the other. In 2026, two philosophies of software production face off. Which one ships software that lasts?
Six layers of explicit, structured, verifiable representation. Where an LLM alone has a representation hidden in its weights, Swoft agents know exactly what they know, and can prove it.
The Sheridan framework adapted to 2026 business: five tiers from simple assistant to fully autonomous system. Diagnosis, architectural requirements and a realistic timeline to move up a level.
Zapier, n8n and Make have dressed their workflows as AI agents since 2024. Seven technical criteria separate a real autonomous agent from automation in disguise.
How to combine LLMs and formal reasoning to build reliable AI agents on specialized domains. A reading of Kautz's taxonomy and the AlphaGeometry, Plato-3, FAOS cases.
Technical diagnosis of the four structural obstacles that block the transition from demo to real production. And what it takes to clear them.
The Model Context Protocol published by Anthropic in 2024 has become the de facto standard to connect LLMs to tools. Why it matters, how to adopt it, what to watch out for.
The Cemri 2025 paper identified 14 recurring failure modes in multi-agent systems. Diagnosis and three architecture families to guard against them.
Generic SaaS has had its day. The new frontier: encoding the actual business expertise, organizational specifics and frictions to erase into the software itself. Software becomes the architecture of the company.
Our R&D team publishes regularly on AI agents, software architecture and Deep Tech. If an angle interests you, drop us a line.