One of my current shipping strategies: Get AI to run a cron loop to check monitoring/metrics/traces every 5 minutes and provide recommendations of further tweaks to make after I've shipped a perf fix So instead of manually verifying the win, I can automate it 馃
I rarely use monitoring/tracing provider dashboards and UIs anymore Only to get graphs to post on here Everything is connected to my agent via MCP, it's so good I'd possibly even go as far as to say all these providers could delete most of their UI and just go agent-native

Oct 2, 2026 路 3:50 PM UTC

13
8
199
15,274
Sort replies: Relevant Recent Liked
Replying to @DanielLockyer
How鈥檚 the noise to useful insight ratio on these alerts?
1
1
562
It's usually very targeted, so high signal For example, "monitor /api/v1/foo endpoint latency every 5 minutes after shipping PR 2323 to verify the improvement"
1
7
1,047
Replying to @DanielLockyer
One of my client companies banned ai usage for metrics in Grafana/Victoria 鈽狅笍 Stone age
1
246
Replying to @DanielLockyer
What do you use for monitoring and traces?
64
Replying to @DanielLockyer
I was doing that without having a proper setup for this... Was basically prompting every day to check what has happened in the previous day. Do you have any system for that or is this a minimal thing?
29
Replying to @DanielLockyer
does the cron retire once the fix is confirmed, or is it still checking last month鈥檚 wins?
2
Replying to @DanielLockyer
Cron loop for metrics is a good idea
20
Replying to @DanielLockyer
Feeding it the pre-fix p95 as a baseline in the prompt keeps those recommendations concrete instead of generic tuning advice.
116
Replying to @DanielLockyer
yeah the same helped me in decreasing api costs by half
46
Replying to @DanielLockyer
yeah, they will easily win there
1
53
Replying to @DanielLockyer
It's a great idea
Replying to @samlambert
we have opencode run every 5 minutes and look at everything from planetscales mcp and tell us about anything problematic
106