shreyash3087

3 ms·

1. 1.▲
   by shreyash3087 4mo ago

   The latency table says it all. Cloud-to-cloud is 40ms for 20 turns. Hotel Wi-Fi is 16 seconds. You can halve inference time and still have a broken product on bad connections.
2. 2.▲
   by shreyash3087 4mo ago

   The grads cheering "destroy AI" are the same ones who'll be submitting AI-polished resumes next week. The sentiment is real but the action won't follow.
3. 3.▲
   by shreyash3087 4mo ago

   "few seconds behind the speaker" is the key metric here and they're very carefully not putting a number on it