@VasuNallasamy@junxi_song as someone who has practice interviewing with voice agents, I feel they're not intelligent enough to probe for depth of knowledge
@Radipdegen@turbopuffer If you’re indexing code from a GitHub repo, you can make a new branch for every commit on the main trunk. This is a way easier model for versioning data.
@akshay_pachaar What results did you get on BIRD? You can easily get 99.7% recall with vector search alone. BIRD tests reasoning for the query composition more than retrieval IMO. Graph databases are a meme
@DamianBitFlippr @computesdk I don’t think it is. I was just working on benchmarking Daytona and 2s is for custom images. Daytona has prewarmed default images that can be created in .3s. I tested this just now. Their CEO says they can keep our custom images warm so they can also have 10x faster creations.