Shared Load
What I notice when systems do part of the work with me
I said I’d share a bit about the tools I use in my day‑to‑day data work. Not a tutorial. Not a deep dive. Just a glimpse into how my relationship with data shifts when the interface starts supporting how I already think.
What I ended up with were four short videos. Individually, they could look like how‑to content. Together, they tell a different story. One about how tools can either force you into their shape, or quietly adapt to yours.
This isn’t about mastering Kusto. It’s about the tools working the way I work. (Also, this was my first time recording videos like this. Please be kind.)
In the first video, I’m asking questions of a Kusto cluster through the GitHub Copilot agent. There’s no query syntax involved. I’m asking in plain language, trying to understand what’s there and how things fit together.
The agent tries. It explores a few interpretations. It reasons about what I might be looking for. But it never quite gets to an answer.
What it can offer instead is a handoff. Here’s the syntax you could run yourself. Here’s how you might ask the system more precisely.
That’s not a failure. It’s a boundary.
The agent can help translate curiosity into something executable, but it can’t yet stay with me all the way to understanding. At some point, it has to step aside and ask me to do the grounding work myself.
The question is easy to ask. The work of answering it is deferred back to me.
In the second video, I install Azure MCP and ask the same question again.
My intent hasn’t changed. What changes is that the system can now carry the question through. Instead of stopping at suggestion, it can execute. Instead of offering syntax, it can return results.
That collapse of distance matters.
I don’t have to switch mental modes from asking to doing. I can stay in inquiry while still being anchored in the data itself. The tool is carrying a bit of the mechanical load, which leaves me more room to think.
I’m able to move more quickly, and more comfortably, into the parts of the work that actually require judgment.
The third video is where the work starts to feel comfortable.
I generate a relationship diagram. This is how I work when I’m trying to understand a system. I sketch relationships on paper as I discover them, adjusting as I learn more.
This is usually a private, manual step. Something I do off to the side while queries run.
Here, the system meets me there.
Azure MCP, working with GitHub Copilot, is doing the same inference I do manually, but much more quickly. It’s pulling together signals, testing associations, and offering a first pass at how things might relate. What it gives me back isn’t an answer. It’s a starting point.
The diagram is editable. I can change it as I learn more. I can question it, refine it, and correct it. But I’m no longer starting from a blank page or a half‑remembered sketch. Some of the initial load has been lifted, which lets me dive deeper sooner and more comfortably.
That matters more to me than correctness. It gives me something to think with, and it lowers the friction of getting started.
Instead of flattening a complex space into a series of answers, the system helps me see the shape of it early, and then invites me to keep questioning. The data hasn’t changed. The work just looks more like the way I already reason.
Why this matters to me
I’ve said before that my professional writing isn’t about how‑to. That’s still true. What I care about is how systems support, or interrupt, the way experienced people already work.
These videos capture a small but meaningful progression. The same question asked with and without execution in the loop, followed by what becomes possible once the system starts sharing some of the cognitive load.
This isn’t about GitHub Copilot, Azure MCP, or Kusto specifically. It’s about the relief of tools that carry a bit of the work so I can spend more of my energy on understanding.
I notice when that happens. Those moments tend to compound. They’re often where the real work starts.
A brief note on the environment
A few people asked what environment I was using when they noticed Kusto Explorer open in VS Code. That’s what the fourth video covers.
It’s included for context, not because it’s the point. The point is not which tools are installed. It’s how little they matter once the environment stops fighting your habits.