[thisisit]

They keep running experiments like free $50 in extra use credits or 2x usage outside certain windows where inference is very slow. You can’t help but think this is all a slowly boiling the frog experiment. Experimenting how much they can charge.

[blharr]

They're boiling the frog pretty quickly, honestly. The token usage has clearly been an issue since using Claude code from the beginning. It just blows through tokens