Skip to main content

Introducing Our AI-Assisted Development Docs 🤖

· 3 min read
Gergely Sipos
Gergely Sipos
Frontend Architect

AI tooling is moving fast — fast enough that "just Google it" no longer works. We just shipped a dedicated AI-Assisted Development section to the Web Hub so the whole team is working from the same playbook.

Why a whole section?​

Six months ago, AI tooling meant autocomplete. Today it means IDE assistants, chat models, autonomous coding agents, and rapid-prototyping tools that can scaffold a full app from a sentence. The surface area is huge and the quality gap between using these tools well and using them poorly is enormous.

Rather than letting everyone figure it out independently, we mapped out the landscape and wrote down what we've learned.

What's covered​

Seven pages, from the big picture down to the practical details:

  • Overview — the full tool landscape: IDE assistants (Copilot, Cursor, Windsurf), chat models (Claude, ChatGPT, Gemini), coding agents, and rapid-prototyping tools like bolt.new and v0.dev. Explains the spectrum from autocomplete all the way to multi-agent pipelines.
  • Guidelines — team-wide rules for responsible use: code review expectations, security hygiene, and knowing when to slow down.
  • Prompt Engineering for Code — how to write prompts that actually work, workspace instruction files, and tool-specific techniques.
  • Vibe Coding — what it is, which tools enable it, and when it's the right call versus when it isn't.
  • AI Coding Agents — how agents work under the hood and practical tips for getting the most out of them.
  • Multi-Agent Orchestration — patterns and frameworks for coordinating multiple agents on complex, long-horizon tasks.
  • MCP Servers — the Model Context Protocol explained: what it is, the most useful reference servers, and how to configure them in VS Code and Cursor.

Our current recommendations​

Two agents stand out. Copilot Agent Mode is the recommended default — it's already in your editor and handles the day-to-day well. Claude Code is what we reach for on complex or large-scale tasks where a deeper reasoning loop pays off.

On the broader spectrum, the docs draw a clear line between AI-assisted development (you drive, the AI assists) and vibe coding (the AI drives, you steer). Both have their place. Neither removes the fundamental rule: you remain accountable for every line of code that ships.

Start here​

The AI-Assisted Development overview is the best entry point — it maps the whole landscape before you dive into any individual topic. 🤖