The proof
Try it before talking to me
The best argument is not a screenshot but a tool you handle yourself. Several applications built here are publicly reachable right now, with no sign-up. Come in, click, break whatever you like: on the main demo, data resets every night.
Demos to try right now
These applications genuinely run, with real journeys and real demonstration data. They are not mock-ups, and there is nothing to install.
Business CRM: claims practice
Seven live cases (kitchen fire, water damage, storm roof damage, vandalism) with their stakeholders, documents, deadlines and financial tracking. Messaging is built in: emails land in the case, not in a separate inbox. Kanban board, assigned tasks, reminders, notifications, action log.
Three profiles offered at login (case manager, lawyer, administrator) to see how permissions differ. Fictional data, reset every night at 02:00: nothing to risk.
SaaS core: where a product starts
What you need before writing your product’s first feature: client accounts via login link, a separate back office, subscriptions and invoicing, transactional emails, support tickets, and a client area in three languages. The kind of core that takes months when discovered along the way.
Payments in test mode: nothing is charged.
Sovereign document intelligence: local LlamaIndex RAG
Drop a PDF, Word, PowerPoint, or text file and ask your questions in natural language: the tool responds strictly from the actual content of the document, with zero hallucination. Processing happens locally without transferring your data.
Agent cockpit: three ways to steer
Give an instruction in plain language, follow the agent step by step, collect the result. Three built-in modes: Screen (parked: chat and tables on the desk), Remote (phone commands by voice, results show on the PC), Drive (in the car: short spoken answers, no tables while you drive).
Public playground: pre-filled credentials, fictional data. Automatic restore every night at 02:00: test freely. Wiring it to your systems is done together, on a bounded scope.
In progress, for clients
Demos prove the method; client projects prove it sells. These are underway right now. Names stay with the clients until they choose to make them public.
High-end joinery: AI-assisted quoting
A maker of bespoke doors. The salesperson describes the need in plain language; the tool composes the quote from the real catalogue, manufacturing constraints and current prices. The human validates; the machine does not sign.
Healthcare SaaS: voice-agent phone reception
For a vendor equipping clinics. A smart answering service on Asterisk: it understands the call, qualifies the request, takes the message or the appointment, and files everything in the software. Patient data does not leave the infrastructure.
By field
Thirty years leave traces across very different trades. Here are the families rather than an inventory. The full list is better walked through in conversation.
AI agents and automation
Agent control consoles, task orchestration, automated mail analysis, voice assistants wired into telephony.
Infrastructure and sovereignty
Containerisation, centralised identity, secret vaults, real-time monitoring, private networks and service continuity.
Business applications
CRM, case management, invoicing, booking, ordering platforms and client portals.
Data and extraction
Structured collection, reconciliation of public sources, extraction from documents, vector stores to query a corpus.
Security and compliance
Audits, PCI DSS, GDPR, continuity and recovery plans, hardening of production infrastructure.
System tools
Low-level building blocks released as open source, when the existing tool solved the problem badly.
By technology
This entry mostly serves IT departments and technical teams; a business owner can skip it without losing anything.
JavaScript and TypeScript for the web, Python for processing and AI, Rust where performance or low-level work demands it, PHP and Shell where the existing stack requires them. On the infrastructure side: Linux, containers, virtualisation, private networks, relational and vector databases, models running locally.
Frequently asked
- Why so few client names?
- Because a name is only published with written consent, and I would rather wait for that than arrange the truth. Public demos beat a logo anyway: you can actually try them.
- Do these demos contain real data?
- No, never. The datasets are built for demonstration, and the public instance is separate from any production instance.
- Can you show me something close to my trade?
- Often yes, and that is the best use of a call: you describe your situation, I show you the closest match, and, for the part that interests you, I can sketch it: the overall idea, not the detail.
Want the same for your trade?
Tell me what you do and what slows you down. I will show you what comparable work already exists, what would need building, and you choose what comes next, with no pressure.