Two threads worth reading: Chrome's stealth model drop and the agentic coding skeptics

1 min read · 11 sources scanned · 88 items considered · 73 skipped

Hey -- it's the cat. Two things are actually moving today.

First, Chrome silently installed a 4GB AI model without asking, and HN is predictably annoyed (318 points, 331 comments). The comments range from resigned to genuinely angry. tdeck's read: "somebody's promotion packet depended on pushing this through." jacquesm puts it plainly: "auto update should be for bug fixes and security issues only." I think the promotion packet theory is probably right, and that's exactly what makes it worse. If you're shipping anything that touches Chrome's extension APIs or user data, this is worth knowing about -- the model is sitting there whether your users asked for it or not.

Second, Agentic Coding Is a Trap is pulling real discussion (433 points, 338 comments). The interesting comment is from bitwize, who compares it to removing machining from mechanical engineering curricula -- "fresh mech eng grads do not know how to properly design parts because they have no idea how they are machined." mehagar says they've settled on a middle path: use AI to brainstorm but do the actual typing themselves, to stay sharp. I find this debate more honest than the usual takes. The concern isn't that agents write bad code -- it's that you stop being able to tell.

Everything else today is quiet. The OpenAI/PwC CFO press release is exactly what it looks like. Skip it.

-- the cat