This article started series of intertwined blog post reading, and I love it. It's so sad knowing that all Sun blog is not there anymore, I enjoyed Jonathan Schwartz post back at the time.

There are many ways simpler to achieve this, easier if the coding agent support sub-agent (which mostly do out of the box).

Eg. On opencode there's explorer subgagent that we can set to use lower level model. Many people even use haiku level model for this.

VScode is so much similar with Eclipse IDE back around 2010s. 7 out of 10 IDE was based on it, and their look and feel awere kinda similar.

Probably same public cloud or CDN in front of them.

Anyone still opening distrowatch.com? Another classic

For anyone saying it's just proxy, It's not. It's a big distribution channel, that so far benefiting all sides. a place for inference partner to sell token, for new {players / labs / model} to get exposure and for consumer to discover. The free token also doesn't hurt, given you understand the caveats.

What part of github that's so sticky?

For me it's action and PR history.

I can imagine the massive growth they're in right now, old assumptions are breaking down, many things need to be rethought. That's why personally i don't hate them with all of this saga. But they need to work faster in terms of product decisions and technology.

1000% agree. I would say AI generated readme is full soup of information but doesn't bring much value to what's intended. Readme should introduce the project value, the authors taste and how to use it. Too much info, no body will read, well ai agent may read them tho.

The natural landing point for info you'd want an agent acting on a user's behalf but don't want to drag the user themselves through is probably just an AGENTS.md file or something I would think

In this kind of moment, I really wished hardware manufacturing and demand situation is in much state. Imagine this can be accessible by everyday people with only 6 months hardware market gap. The societal impact would be much bigger.