RUNS ON
Lucid AI Labs is run by Fabi, a solo founder in Germany who builds and operates every system personally. That is not a limitation we apologize for. It is the product.
You are not customer number 4,000 in a queue. There is one pair of hands on your system, and you know whose.
The honest version: this studio began as a tool to structure my own days for my first company. Nothing more. What it turned into was not planned.
I needed something to structure my days for Essential Elements, my first company. Nothing ambitious. A side project. That is the whole reason any of this exists.
Piece by piece it grew into Lucid, my own AI system: a persistent memory graph, semantic vectors, autonomous context management. Two weeks in, the first custom model training was already running.
Not in a bad way. Even a small version of Lucid showed an independence I had not expected. From then on this was not building tools. It was the foundation for everything that came after.
Once I started seriously testing the AI products on the market, the pattern was hard to miss: cheap models, generic prompts, no escalation paths, no transparency. Lucid AI Labs exists so the systems we ship are not one more of them.
Generic prompts on cheap models, no understanding of the customer's business, no safeguards. That is why AI already has a bad reputation in many industries before it has properly arrived. Someone should be doing this right.
When I build AI into your business, it is not at the quality you see everywhere else. It is at the quality I built for my own system.
Fabi, founder
Fabi. Best described as an amateur-level AI researcher: nine months of custom model training, a self-hosted AI system that runs around the clock, and an honest opinion on what works and what does not. Not a salesperson.
Small means no account manager, no sales engineer, no junior passing your brief along. You talk to the person who builds the system, operates it, and checks the server at 10pm if something breaks.
Few projects at a time, chosen carefully. When the work does not meet the standard, it does not ship. That rule only survives if the project list stays short. The cap is not a growth problem, it is the quality standard.
Registered German business. Systems self-hosted in Germany where the product calls for it, DSGVO handled as an engineering requirement, not a footnote at the end of a contract.
A voice line you can call, a report you can read, systems that answer in public. Everything this site claims can be checked before you spend a cent. Artifacts, not adjectives.
No sales team, no ticket queue. Write what problem you want gone and you get a straight answer, usually the same day, from the person who will build it: [email protected]
The personal AI system that started all of this is still running, and it is still the proving ground. The same inference server, memory layer and data residency carry into everything the studio ships, from Verdict reports to chatbots and voice agents.
If you want to see who is actually building your system, look at the repos. No marketing repo, the real thing.
A personal body-state engine. A thin iOS client, a correlation server and a DIY BLE bridge. Every algorithm runs server-side.
SwiftBody-state-aware motorcycle ride telemetry. A SwiftUI app, heart data joined from Supabase, telemetry tiles for lean angle and GPS.
SwiftAn MCP server for Uiverse.io. Browse, search and drop in 3,800+ open-source UI components from Claude Code, Cursor or any MCP agent.
TypeScript