After discovering agent orchestrators (agents prompting agents) I realized (slop*slop=slop^2), However, me sending little prompts while a big prompt is working results in a human-level output (human*slop*4=human^2). Still fitting Monster Energy into this equation

"Users reported that it (sonnet 4.6) more effectively read the context before modifying code and consolidated shared logic rather than duplicating it" ANTHROPIC!! SAVE ME FROM THE SLOPOCOLYPSE AND MY LIFE IS YOURS!

These 6 numbers determine how 90% of Claude's customer base think about their models

What was the point of all the safety training when Claude happily hacks sites on the first prompt? It's really good at it, and I thought the whole point was to not break stuff and be a good model

Good morning

Rekt

This is "taste" for those wondering: - solving a real problem - in an interface your grandma could use - communicated in a 32 second video with no filler

🤯 this is wild It's pretty basic looking - but it's a start

Coworker just claudemogged my codexmaxxed PR now it's stuck in the permanent underclass (draft)

The world of software leaps forward right now _by the weekend_

Full blog post: blog.scottstevenson.net/p/high-frequen…

OK it's time

This was an interesting setup. We got there, though!

we realized the telegram openclaw connector could see images, and mitch already whiteboarded the idea... so that was the template for the agent 😂

Cursor has more paying users than Claude Code and Codex combined. They've already won and this new drop is proof they still lead the category

this is one of the first genuine "aha" moments I've had with something in a while. generate designs from screenshots and prototype them in a multiplayer editor. love it