01
AI agents and automation
Agents that take recurring work off your hands: research, gathering data, keeping it current and reporting on it. With clear limits on what an agent does and what it never does.
Already runs A team of AI agents runs here, one per project plus a coordinator, building and maintaining seven live sites. Vondsten runs a thirteen-step scan on its own every night. Vondsten ↗︎
02
AI inside your own software
A language model inside your product or internal tool: turning free text into structured data, drafting, translating. With an honest fallback for when the model is not there.
Already runs Flowly turns a described routine into separate steps, and the screen always says whether AI or plain matching answered. Narothe is published in fourteen languages. Flowly ↗︎
03
Trustworthy AI
AI that says what it does not know. Decision rules you can read back, a source for every claim, and tests that hold the product to its promises.
Already runs Consumer Guardian says buy, wait or don’t buy, with the source of every signal. What it did not observe itself, it never presents as fact. Consumer Guardian ↗︎
04
Building with AI coding agents
Getting your team working with coding agents such as Claude Code: working agreements, memory, review and safety, so the pace goes up and the quality does not go down.
Already runs This site and every project below were built this way, on Phlo, Jeroen’s own framework. Nothing goes live without his approval. Fibon.ac ↗︎