At some point soon, we're going to have to admit that we no longer truly understand the systems we're building. Companies are openly bragging that their developers don't bother looking at the code their agents produce; and it's so much of it, that code reviews have become a passive thumbs up. Even without AI generated features, modern systems are already extremely complex which makes it challenging to debug, especially in areas where you're not a subject matter expert. Now is not the time to slack on observability. If anything, the need just went up. We have to be able to properly investigate incidents when they inevitably occur. I like Honeycomb Canvas for this because it gives me an interactive workspace where I can explore and debug in natural language without manually piecing together queries across telemetry data. "Why was the checkout flow so slow yesterday?" "Customers are complaining about site crashes. Help me figure out what's wrong." "There are reports of an outage... how widespread is this?" Canvas runs queries on your traces and structured events, giving you visualizations to help you quickly get to the root of the issue. It'll even suggest follow ups if you're not sure where to look next. My favorite usage is @'ing the Canvas Slack bot to pull insights right into team discussions 👌🏽 If you're not already using Canvas, it's definitely worth checking out: fandf.co/4djGo6k Shoutout to the good folks @honeycombio for the collaboration. Really enjoying Canvas, and I've been playing around with the MCP server too :)