It’s always been a challenge for me to maintain a mental model of the product.

AI and agentic development is accelerating the problem of maintaining a systems model that allows us to make good design decisions. The more code we create, the more we need to review and learn.

The long poles in the software development lifecycle that have a lot of room for improvement are comprehension & alignment.

• How does it work now?

• How should it work?

• Where should we go next?

• How do we get there?

I think it’s becoming abundantly clear that commoditizing coding is not commoditizing software development.

Stepping aside to speak to my own interests —

I don’t want to lose touch with how it works. I feel a lot of builders love solving the puzzles. When I say that, I’m not talking about writing code, I’m talking about shaping the solution, testing the solution, and improving it.

The loop of shaping the product is my joy.

In the end, we can all benefit with a hand making code more accessible — whether it’s the code you forgot you wrote last week, your colleagues PR, an open source repo or the Agent who just built something while you got coffee.

My solution is called Intraview.

It’s a VS Code extension (Yes, forks like Cursor too) that works with your existing Agent (Claude, GPT5-codex) via a local MCP to build visual tours. The tours can act as guided feedback sessions, as well, where you can review some work by an Agent or a PR with added context by the Agent on how things work. It’s secure-by-design, with the only external calls being some basic anonymous usage telemetry to Azure logging to help me understand what features to prioritize.

Nevertheless, it’s better to show than tell.

I created a walkthrough where I show how I pull down an Open Source repo (Plotly JS), install Intraview, create a code Tour to get me started with learning how to build a new visualization.

I also show how the tool helps with giving agents feedback.

It’s not a polished marketing video, there is no flashy hook, I assume you’ve read this and I don’t lead with why. It’s pretty raw outside of some speed-ups to spare you from my bad decision to choose a Repo that’s 1.5GB.

The scene opens as all great stories with a Terminal and a GitHub page.

I’m grateful that you’ve come this far to engage with this work and hope you’ll take the next step to help me shape it.

Let me help you build and enjoy it a little more than you did yesterday.

Enjoy and, please, tell me what you think on HackerNews.

— Cy

If you’re interested in following my progress, find me on LinkedIn (I don’t use any other networks).