Engineers finally get to work on the system
When an agent gets it wrong, the fix is a rule, a skill or a hook rather than an edit to the code, so the engineer's job moves from working inside the system to working on it. Engineers were rarely given that job and have few criteria for doing it well, so a team has to decide what the work is judged by.
Written by
-
Benedikt Stemmildt Founder & Co-CEO
Published: 23 September 2026 Updated: 1 min read
I don’t review my coding agent’s code line by line. I look for the rule it was missing.
Like pairing with a new colleague
It works like pairing with a new colleague. You watch how they work, see where they keep going the wrong way, and tell them.
A harness does not remember what you tell it. Each correction has to become a rule, a skill, or a hook that fires a check whether the agent thinks of it or not.
Turning corrections into rules
At the end of a session I send the agent back through our conversation and the ones before it. It looks for the corrections I made that it did not think of itself.
Then I ask what would make it follow them from now on. It comes back with new rules, a skill, sometimes a scheduled check.
So I work far more on the system than inside it. The architecture rules are what I still write by hand, and I check each hook myself.
I also run a skill that goes through past sessions and updates the rules on its own. I am not sure yet that it helps, or whether it only writes down what the model knew anyway.
A job engineers were rarely given
Engineers were rarely given this job. For years we sat deep inside the system, building for everyone else, with no time left to work on it.
That is why we have few criteria for doing it well. Architecture shows the same gap: we know we want software that is secure and easy to change, and rarely say how secure.
Rules, skills and hooks are now where a team writes down how it works. Make building them someone’s job, and decide what that work will be judged by before the first rule goes in.
Further reading
Written by
-
Benedikt Stemmildt · Founder & Co-CEO
More on Adoption
- Five human skills that grow more valuable with AI My son asked why he should learn programming if AI will do it for him. The skills that matter most are becoming more valuable: critical thinking, empathy, language, value creation, judgment and taste.
- Give every team one AI hypothesis instead of generic training Generic AI training across 300 people produces compliance, not adoption. One written hypothesis per team, tested on a retro cadence, makes the rollout correct itself.
- Find where the work piles up before you pick an AI playbook Four engineering organizations put AI on four different problems. What transfers from them is the diagnostic they all ran first: where does the work actually pile up?
Scaling needs named owners
Coding agents scale across teams when named people own the platform, the team practice, the guardrails, and the measurement. Rollout & scale is where that ownership is built and handed to internal owners, in your stack.
Training on this topic: Agentic Engineering for Engineering and Product LeadersWant to check one team first? The self-check: twelve yes or no questions, three minutes, instant result, no account. Take the self-check