How on earth did people get anything done before agents?
热度趋势
趋势数据积累中
百分比基于当前可用热度信号，而非评论数或独立用户人数。
一位开发者表示，在使用代理后效率提高了10倍，目前正在寻找模型路由的替代方案。他们正在使用openrouter构建自己的模型路由器，但效果不佳。他们欣赏standardcompute.com的模型路由器，因为它采用月度订阅制且没有随机不可预测的5小时使用限制，但指出它不支持免费模型。这位开发者正在寻找其他好的、最好是由风险投资补贴的替代方案。
I have personally become 10x more effective compared to one year ago. I love it as there is always so much I want to do!
Right now my rate limiting factor is usage limits by far. The token bill bites quite hard and make it hard to justify some projects, so I’m on the lookout to maximise agentic work per dollar.
Codex/claude etc. is subsidised atm, but they are also wasting a lot of tokens on expensive, non-open source models that could have done the job way more efficient.
Therefore been building my own model router with openrouter but not working very well (https://openrouter.ai/docs/guides/routing/routers/auto-router ). Think standardcompute.com has an excellent model router and like that it’s a monthly thing and not random unpredictable 5-hours usage limits etc. However, they don’t serve free models which would be nice.
Any other good alternatives right now? Preferably heavily subsidised by VC money 💰