The Harness Came — Now We Are on the Loop

Part 3 of our Claude Code journey: engineered environments that check the model's work, and the loop — plan, act, verify, repeat — that now ships our software.

busoft.co.il

Part 3 of 3 — the Busoft Claude Code journey.

Once the context was engineered, the question changed again. A model that can work for hours raises the stakes: who watches the work? Our answer — the industry's answer — was not more supervision. It was engineering. The harness came.

What a harness is

The harness is everything we build around the model so that autonomy is safe: permission boundaries that decide what it may touch; hooks and gates that run checks after every change; skills that encode our procedures — how we package a release, how we review a schema change; sub-agents for isolated work; memory that persists what matters. The model stopped being a chat window and became a worker inside an environment that constrains, verifies and empowers it. The pattern is now well documented — from Anthropic's Claude Code best practices to retrospectives on four years of agentic patterns — and it is the layer where most of the engineering value now lives.

Now we are on the loop

With a harness in place, the last step is natural: close the circle. Plan, act, verify against evidence, and go again — until the goal is met, not until the output looks plausible. This very website is the proof we show customers: it was built, packaged and deployed by agentic loops running inside a harness — a deployment script that checks the database before touching it, verifies its own configuration after writing it, streams the application's logs while it warms up, and prints its own diagnosis when something fails. When the site went live, a watching loop verified every section in both languages and notified the search engines. Nobody babysat it.

What this means for your team

The engineer's job is moving up a level: from writing the code to designing the loop that writes, checks and ships the code — and proving it with evidence. That is an organizational change as much as a technical one, and it is learnable. It is exactly what we teach in the Busoft Academy, what our consultants install in delivery teams, and what we build for customers end to end. The era has a clear arc — prompts, then context, then the harness. We are on the loop now. Come join us there.

Read next

Latest from the blog

Back to blog
busoft.co.il

Busoft Team

The New AI Architecture Era

Software is entering a new architecture era: self-improving development loops, AI teams designed like organizations, and a strict boundary between reasoning and execution. Notes from our internal research.

busoft.co.il

Busoft Team

From Prompts to Context Engineering

Part 2 of our Claude Code journey: the moment we stopped polishing sentences and started engineering what the model sees — and what it did to our delivery speed.

busoft.co.il

Busoft Team

How We Started with Prompt Engineering

Part 1 of our Claude Code journey: what prompt engineering gave a 25-year-old software house, where it broke, and the lesson that pushed us forward.