What’s shocking about it? Seems like the usual culprit— a bad config rollout. Took a long time to identify, so maybe that’s shocking. But I can attest that sometimes, you get into fight or flight mode and miss the obvious when trying to diagnose a disruption like this.
That said, nowadays, the first thing I do is spawn an agent to look through the most recent commits and try to identify something that could be the cause of a service outage.
This one seems like something Claude Code or Codex would have quickly flagged.
Agreed, we've all been there, but 4 hours! For a network config change. No one raised their hand and said "hey I just toggled this thing maybe we should look, I did it exactly when our entire region went had down"
Has any harness matched the effectiveness of Claude Code yet? I haven't experimented much recently, but every time I have in the past, I wasn't able to get any other tool to approach how effective I am in CC.
I'd love to use a different harness-- ideally an OSS one-- and hook it up to whichever LLM provides the best bang for the buck rather than being tied to Claude.
OpenCode has been great in my experience. I still get the best results using it with Anthropic's models, but some of the open weights ones are catching up (GLM 5 works reasonably well for me).
Some of us think that a key aspect of society is that we take care of each other. If something terrible happens to you before you manage to amass a fortune, it’s nice to live in a society that won’t leave your family destitute.
True, but reductio ad absurdum is a good way to make any argument look silly without actually considering nuance. Of course, there's some limit to what society will do to save an individual. If someone is lost at sea, we'll try to save them, but we won't spend $1T rerouting all of our available naval capabilities to do it. How much should we spend? The math isn't clear, and thus the economics aren't clear. But, where we should fall is somewhere on a gradient between "Every man for himself" and "Save every individual at all costs."
The question is, where do we fall on that gradient?
Some of us don't like paying for other people who make objectively bad decisions that cause them to need to be bailed out in some way.
There's nothing wrong with taking care of others, but there has to be limits. Hopefully the limits are designed in ways that encourage objectively good choices and discourage objectively bad ones.
I’m using it in a >200kloc codebase successfully, too. I think a key is to work in a properly modular codebase so it can focus on the correct changes and ignore unrelated stuff.
That said, I do catch it doing some of the stuff the OP mentioned— particularly leaving “backwards compatibility” stuff in place. But really, all of the stuff he mentions, I’ve experienced if I’ve given it an overly broad mandate.
Same. This weekend, I built a Flutter app and a Wails app just to compare the two. Would have never done either on my own due to the up front boilerplate— and not knowing (nor really wishing to know) Dart.
reply