heyjstn

it's an arm race, everyone know exactly what that would lead to, including people in the frontier labs; but the mindset is "if it's not you then other will do, so better you're the first comer to the party so you can still get the best part of the cake"

^this, i know the market is bad but the author seems like *whining* a lot. i've viewed his CV and the books but they are not as impressive as he described. i'm a software engineer and seeing that software engineering as a professional career have been changed forever

if you're still thinking like that, better switch your career now

I think the next models will be benchmaxxing on the Pelican benchmark tbh

doom marketing at its finest

Have anyone tried a workflow that:

- Fable 5.1 for planning/adversarial reviewer

- Opus 5.5 for well-scoped tasks break down

- Sonnet 5.5 for these well-scoped tasks implementation

I think the blocker might be how efficient the context is compacted and sending around between these agents

I'm tired of reading the Claude-ish generated visualizations, it's fine that you're using AI to write the post but at least please trying to polishing the article if you really care

Others must slow down, but not us.

How could the flash model beat the pro on the Cyber benchmark?

Why the website and the images look like AI-generated?

are you working at anthropic? you're literally repeat what that freaking Dario say everyday