On Claude getting too myopic when working with large codebases

“A skillful driver can mitigate this, but most lack the skill”

i'm referencing a post that can be found here: https://x.com/steipete/status/2012346819372532094?s=12

“Just two missing characters in a Regex filter allowed unauthenticated attackers to infiltrate the build environment and leak privileged credentials”

This is a well written disclosure document 👍

https://www.wiz.io/blog/wiz-research-codebreach-vulnerability-aws-codebuild

I just installed opencode using the “oh-my-opencode” wrapper and asked it to add a feature which I wanted to build in my side project, and guess what… in just 2 big shots, it was able to make a working version of it with minor tweaks from my side.

The value add is insane!

It genuinely feels to me like GPT-5.2 and Opus 4.5 in November represent an inflection point - one of those moments where the models get incrementally better in a way that tips across an invisible capability line where suddenly a whole bunch of much harder coding problems open up

It's possible Gemini 3 Pro should be included in that group as well, but I'm not seeing quite the same level of astonished buzz from hardened software engineers around that model as I am for the other two .

sourced via x

I was able to close 30+ junk pull requests in essentially no time!

I wrote about it here: https://tnvmadhav.me/til/closing-github-prs-in-bulk/

I am currently unable to add images to GitHub issues :(

Is this an outage?

https://www.githubstatus.com says otherwise

Posted about it on X

https://x.com/TnvMadhav/status/2006248437457850703?s=20

Ah, got it. The image file was too big (>10MB)

I tried to add the file in a pull request and the error response told me this.

All good. Perhaps the issue was that the error response in @GitHub actions is different from pull requests.

We are at a point where we don’t usually care whether content is pure / semi ai slop anymore because the prompters are so good at emulating content that we deem as real and this means, another cambrian explosion of content is imminent.

This is just 3 years since the first milestone.

Technical rumination was meant to be but emotional rumination wasn’t.

Quoting BrendanGregg comment on his post 'Leaving Intel'

I like to measure things. In real life and on computers. But I also have a couple of work reasons for it:

As a remote worker, I'm under extra pressure to prove that remoteness works.

As a senior employee, I'm also under pressure to regularly report where my time is spent.

reference to the hacker news comment: https://news.ycombinator.com/item?id=46170913

youtube finally added a chat feature