We build our own products and software for companies, with agents and checks. Here we write down what works, what went wrong and how we go about it.
Agentic engineering
One model reading your code finds a lot and misses a lot. A council puts several specialists on the same code, has their findings checked and leaves the decision to a person.
, 6 min read
Agentic engineering
An agent that approves its own work misses its own mistakes. This is how ARCS keeps making and judging apart, and how we decide where a person has the final say.
, 7 min read
Agentic engineering
Vibe coding is fast and fine for a prototype, but nobody reads the code. When that is fine, when it gets risky and how to get from prototype to production.
, 7 min read
Agentic engineering
Building software with AI agents while people set the goal, the limits and the final check. Where the term comes from, how it works and what breaks without checks.
, 10 min read