I have never felt more seen

Guy in the middle doesn't care about the limits the guy on the right now has to solve. They usually pop up during an incident and you're trying to bandaid architectural problems.

Finally found a place to use these giant googly eyes. The disappointed mini-split of shame judges you while you're working out.

One quick note about the Hugging Face incident, don't give your pods too much access. I'd have to understand their sandbox setup but this isn't least privilege and it doesn't sound like they had network policy in place.

How does the AI Software Factory work with IaC or GitOps? I've seen agents make PRs that would delete pivotal AWS resources or critical components from clusters. How do we trust the agents doing code reviews more than the ones contributing code?

At the end of every day I close all my tabs. It's easier to find it all again then figure out what the hell I was doing the day before.

These are the voyages of the Startship Claudiprise, to boldly go where no AI has gone bef... You're right! We've gone too far. My analysis is our warp drive failed and navigation has broken irreparably. Life support systems are at 7% that's not just dangerous, it's deadly.

Sitting in chemo, see awesome Dewalt saw deal @Lowes, order! 5 hours later, "ready for pickup", cool! 1 min later, "order canceled" Try CS in app, kicked from queue. Back in, spot 54. Finally talk, price error. Nothing they can do, recommends I call. "Mylow" so appropriate!

Funny product shot from the new FlipperOne documentation docs.flipper.net/one/general…

I haven't shared this but I was diagnosed with cancer a little over a month ago. Had surgery to remove a tumor. Going through my first round of chemo. I'm still here, putting in the work.

Chuck Norris died! Winter is coming!

I've seen AI agents do dumb things. Everytime someone says "No you have to do ABC, then XYZ" That makes it a tool, not a replacement for the one using the tool. I foresee a lot of engineers getting hired back in the next 24 months.

Be me: Force sync CoreDNS cluster service after upgrade Call at 2AM, one endpoint can't resolve Can't resolve endpoint in any cluster but can locally Enable DNS response logs See "tc" in log entry. Googlefoo Response too big, requires TCP I had deleted TCP from service, kept UDP

Who else misses LimeWire?

One day client/server communication will look totally different. If both have access to the same local model & temp=0, one node doesn't send all data to the other, just the prompt used to generate the information it would have transmitted. Probably 10-50x compression.

7 year olds blogging "tying our shoes was too complex so my team went back to velcro, read why you should too!"

I keep seeing these "Kubernetes was too much, we moved to..." articles and I can't figure out why. One even said they went back to bash scripts. These teams can't have more than a handful of services. Also I am interested in why people think it's hard.

Reading this book I realize now why the em dash is so prevalent in AI output.

Bought some furniture and my oldest sees the invoice. Him: Oh, cool, you know the manager? Me: No Him: How'd you get the manager's discount? Me: Just ask for a discount Him: You can do that?!?! I have failed him.

THROW YOUR XML PROMPTS IN THE TRASH! Instead, use YAML. I promise you, it's better.... PSA: I've never tried either for prompting.