Tools
What we build with, and what we build ourselves.
How we buildTools
Good tools are boring. We choose proven technology and use AI where it really saves time.
Method
How we build with AI
We design the system, the models and the tests. AI agents write most of the code. The responsibility stays with us.
- 01
System
First we design the system: data, processes, limits. On paper, before any code.
- 02
Tests
We define how we see that it works. The tests come before the code.
- 03
Agents
AI agents, above all Claude Code, write a large part of the code. Fast and by our rules.
- 04
Review
Every change goes through review, tests and measurement. Only then does it go live.
Stack
Our technology
No framework of the week. These are the tools we really build with.
WebWebsites and apps
Static pages where possible. Servers only where needed. That makes websites fast and cheap to run.
DataModels and analysis
Forecasts, simulations and tests on data the model has never seen.
AIAgents and assistants
We use AI for code, research and automation. In client projects it answers only from checked sources.
OperationsHosting and versions
Every change is versioned and has a preview before it goes live.
Own tools
Built by us
Tools from our projects and from the team. Some are open on GitHub.
- In useAI Visibility CheckChecks which AI crawlers read a website and what AI systems say about a company. The base for our audit.To the offer
- In QFTForecast engineSix models, a walk forward backtest and 2000 Monte Carlo paths. It calculates the daily budget in QFT.To QFT
- Open sourceBasketball CheckerA board game with its own bot. A linear model rates every move. Tested with 5000 simulated games.Code on GitHub (opens in a new tab)
- Open sourceAssistantHub ProA desk assistant on an ESP32 with AI chat, Telegram bot, web UI and habit tracking. 3368 lines of C++.Code on GitHub (opens in a new tab)
