you gotta use faster models, this is the next big leap in agentic coding. In 2 years we will have opus 4.5 at 1000 tokens/sec and it will be glorious.