First off, the app is beautifully designed and inc...
# feedback
m
First off, the app is beautifully designed and incredibly convenient. However, my biggest piece of feedback is that it burns through context like crazy. For tasks where I normally stay around 220k tokens, I'm seeing usage closer to 500k. I'm essentially seeing double the context consumption, to the point where I have to deliberately manage limits with the Claude models. This high token burn is going to be very tough to justify, especially for models with smaller context windows like Codex. I highly recommend your team benchmark this: what is the actual improvement in output quality versus the context overload? If it costs 2x the context for the same result, I'm not sure the trade-off is worth it right now. That said, the app itself is stunning. I really hope you make progress on optimizing context usage in future updates!