“Prompt engineering focuses on what you say. Context engineering focuses..."

on what the model knows when you say it.” — Dhunay

In V1.0.0, we added something small but meaningful:

a first step toward real team collaboration on the same codebase.

While working on this release, I asked Copilot to help me integrate my commercial codebase with the open-source project. I told it clearly: keep the TypeScript types aligned so local users can switch to Claude without losing any data.

But Copilot did something interesting — and honestly, frustrating.

It assumed it already “knew” both projects. It immediately started making changes, beginning with editing cloudsync. From my inspection, it reviewed my commercial repo but completely ignored the open-source codebase I specifically asked it to integrate with. So I rejected the changes.

Then I turned on context sync.

And that’s where things shifted. With the right context available, the model finally understood the full picture instead of guessing. The difference is visible in the screenshot.

This release is a small reminder of something we tend to overlook:

large language models aren’t just about prompts — they’re about giving the model the right context to reason from.