Welcome back to In General: one useful thing I tried, built, or figured out this week. AI workflows, productivity systems, dev tools, side-project mechanics. All road-tested before they show up here. If someone forwarded this to you, subscribe here.
Glance started because I had a problem.
I had more and more agent jobs running in the background. Cron jobs, monitors, recurring checks, workflow reports, and little bits of automated work scattered across different systems. One or two are easy to track manually. Dozens are not. Hundreds definitely are not.
The easy version would have been a status page.
I could have made a static HTML dashboard that showed a snapshot of activity: what ran, what failed, what looked stale. That would have been useful. It might have solved the immediate problem.
But I did not build it that way.
I started building Glance like a product.
Part of that was practical. I wanted it to live off my machine. If I had different agent jobs running in different places, I wanted a single location where all of them could report their status.
Part of it was less practical.
I wanted the reps.
I wanted to build the product-shaped version to practice making products. And, honestly, I thought it would be fun.
It sounds inefficient. It sounds like the thing you say right before spending three days polishing a settings screen that no one asked for. And I have been guilty of spending time tweaking systems vs. doing actual work.
But AI has changed the cost of that kind of exploration.
Without AI, I probably would not have built Glance. Not because the idea was impossible. It just would have been too much work for something that was not clear yet.
And Glance was not clear.
The first version was going to be a monitor for OpenClaw instances.
Then it began to evolve into an agentic Kanban workflow system.
Then it became what it is now: a centralized monitoring hub for agent work.
That wandering would have been expensive before. Now it was cheap enough to do while I was still figuring out what the thing wanted to be.
That is the part I keep coming back to.
Productizing your work is not just turning an internal tool into an app. It is using product shape to discover what the work actually needs.
A dashboard can answer “what happened?”
A product has to help me decide what to do next.
That difference showed up quickly with Glance. A static page could show me a list of monitoring events. But once I treated Glance like a product, I had to make harder decisions.
What deserves an alert?
What should be visible at a glance?
What context does each alert need?
What is running?
What is it called?
What job was it supposed to do?
How do I remediate the problem?
What was the expectation in the first place?
That last one matters. A monitor without an expectation is mostly noise. It can tell me that something happened, but not whether it was supposed to happen, whether it happened on time, or whether I should care.
The product shape forced those questions into the system.
It was not enough to dump events onto a page. I needed monitor definitions, payload history, health states, acknowledgments, escalations, and enough surrounding context to look at a problem quickly and act quickly.
That is what the static dashboard would not have taught me.
It would have shown activity. Glance made me define attention.
There is an obvious trap here.
Not every internal workflow should become a product. Some things should stay scripts. Some things should stay checklists. Some things should stay one-off commands you run twice and forget.
It is easy to spend all day building tools around the work instead of doing the work.
Productivity porn, but with components.
AI does not remove that trap. It probably makes the trap easier to fall into. When building gets cheaper, it becomes easier to justify building one more layer around the thing you were supposed to do.
The useful change is the feedback loop.
I can MVP the product-shaped version quickly enough to see whether it teaches me anything. If it does, I keep going. If it does not, I should collapse it back into the boring version and get back to work.
That is the test I am starting to use.
Does productizing this workflow quickly create leverage, clarity, or reusable judgment?
With Glance, the answer has been yes so far. Building it as a product helped me define the problem more clearly. It forced me to decide what healthy means, what stale means, what deserves attention, and what context I need before I can trust the signal.
It may become a real product. It may mostly remain the thing I use to manage my own systems.
Either outcome is fine right now.
The point is that the product's shape taught me things the script would not have.
And if it stops doing that, I should stop productizing and ship the boring dashboard.