Rich Snapp

@richsnapp.com

Software Engineer @ MasterControl

now i see the issue. actually wait. the issue is the opposite. now i understand the full picture! actually… let me reconsider. here’s the actual problem: there is no problem.

running your agents both in tmux and inside a container makes it super convenient to clean up all the orphaned processes they tend to leave around. maybe someday they'll clean up after themselves.

htop highlighting orphaned processes

i used windows powertools to map all my mac hotkeys to the windows key. e.g. window + a = ctrl + a window + c = ctrl + c window + v = ctrl + v window + q = alt f4 etc. i should have done this a long time ago.

I am confused by claims that the agentic era is going to mean the end of UIs. Chat really sucks for a lot of use cases! UI is often really great!

I've automated the way I do work with a more structured and formalized ralph loop. now I've written about it for those who are interested in doing something similar, as well as released the code if you want to try it for yourself: www.richsnapp.com/article/2026...

Automating agents with the next iteration of Ralph - RichSnapp.com

If you've been paying attention to the agentic coding space lately, you've probably come across the concept of a "ralph loop." The idea, popularized by Matt Pocock and others, is straightforward: put ...

richsnapp.com

i have an alias i created: clauded='claude --dangerously-skip-permissions' i'll switch to it when i get sick of accepting permissions. it's also a nice verb so when my computer gets fubared people can be like "damn, you got clauded"

claude should track how long its tool commands take to run. maybe then it wouldn't decide to run the same 10 minute test suite 5 times over just to slightly tweak the output format each time.

Agentic engineering is driving the cost of code to zero but the cost of good taste and product sense is going to go sky-high.

so i'm going through the react docs for the latest react 19 features... it's awful. (not the docs, the docs are good) basically a laundry list of caveats, pitfalls, and IOUs. i honestly don't know how anyone could read these and think "i should be using this"

types still exist even if you write javascript instead of typescript, they're just in your head (hopefully) instead of on disk. this is more relevant now than ever because types are some of the most important context you can give AI when writing code.

please do not use llms to actually write your blogs. we can absolutely tell. and it means: 1. I don't trust what you're saying (or at least I don't trust any purported 'insight') 2. I don't feel like I'm listening to you (normally I _like_ listening to nuance) 👉 shorter, imperfect yet real👌