3 ms·
That implies a throughput of around 16 million tokens per second. Since coding agent loops are inherently sequential—you have to wait for the inference to finis
by storystarling 9mo ago
That implies a throughput of around 16 million tokens per second. Since coding agent loops are inherently sequential—you have to wait for the inference to finish before the next step—that volume seems architecturally impossible. You're bound by latency, not just cost.
- mrob 9mo agoThe original post claimed they were "running hundreds of concurrent agents": https://cursor.com/blog/scaling-agents https://cursor.com/blog/scaling-agents
- simonw 9mo agoIt was 2,000 concurrent agents at peak. I'd still be surprised if that added up to "trillions" of tokens. A trillion is a very big number.
- mikkupikku 9mo ago16 million a second across 2000 agents would be 8000 tokens per second per agent. This doesn't seem right to me.
- Snuggly73 9mo agoI mean, its right there in their blog - https://cursor.com/blog/scaling-agents https://cursor.com/blog/scaling-agents "We've deployed trillions of tokens across these agents toward a single goal. The system isn't perfectly efficient, but it's far more effective than we expected."