aiedwardyi

the token split is wild - 93% input vs 4% output. makes sense to optimize output but forcing short responses can hurt coherence in longer agentic sessions

good that they walked it back but the fact it shipped at all says a lot about how these tools are being monetized. trust is hard to rebuild once you start injecting stuff into developer workflows

interesting to see cost-tracker.ts in there. makes you wonder why they track usage internally but don't surface it to users in any meaningful way

This is the core issue. These tools operate with very little transparency about what they're doing under the hood. Even basic stuff like how much of your session resources have been consumed is hidden from you in most tools.

This matches what I'm seeing too. The refactoring and test generation use cases are where it actually saves real time. The tricky part is longer sessions where you lose track of context window and quota usage.

Yeah this is brutal. I started tracking my 5h/7d remaining in the terminal just so I could pace myself and stop getting blindsided. At minimum Anthropic should surface this somewhere in the UI.