> At the time of writing, there have been 244 releases to Claude Code this year; over 3/4 of these changed the system prompt or tool descriptions.

ooof, that is a lot!

And we're supposed to believe them and start one-shotting our projects...

▲

dannyw a day ago | parent [-]

Alternatively, what you’re seeing is continuous hill climbing and optimisations based on large scale experiments and simulated rollouts.

For me personally, I don’t use Claude Code because I can’t read the source.

At least with Codex, I know there’s a bunch of telemetry and all that, but I can have my agent inspect the source and have relatively greater confidence it’s not implanting spyware onto my machine, because a classifier thinks I’m working for Deepseek or something.

▲

lifeisloving a day ago | parent | next [-]

They're hill climbing by changing their system prompt... That's not how thats done.

They're changing system prompts to control how users use their llms.

▲

oblio a day ago | parent | prev [-]

> Alternatively, what you’re seeing is continuous hill climbing and optimisations based on large scale experiments and simulated rollouts.

Let's be real. If their tools were as amazing as they're touting, Claude Code would be a tight, hyper optimized Rust/Zig/FreePascal/whatever 3MB binary that would launch instantly, have no trouble scrolling text at lightning speed on its 20 platforms, etc, etc.

	▲	strangegecko 8 hours ago \| parent [-]
		That's assuming they're optimizing for those things. The software industry hasn't cared about optimization for a long time. All they care about is how to capture profits and market share. The truth is that these optimizations don't do that.