Astra for Law

Commenters picked apart OpenAI’s Astra for Law announcement by focusing on capability, benchmarking, and positioning. Nerevarthelame pointed to Vals AI numbers showing Astra slightly behind Claude and Muse, while smusamashah and others flagged weak scores on other benchmarks. Several commenters read OpenAI’s explicit naming of customers like Harvey and Legora as a strategic signal - piker and jmkd suggested it’s meant to reassure partners or pre-empt competition - while dzonga and cmcclellan argued the company is still searching for a clear go-to-market lane. JumpCrisscross and others criticized the privacy language as vague, and cbg0 and others raised hallucination concerns and called for more disclosure about reliability.

Opinion divides sharply on real-world impact. Halamadrid, droidjj and some practicing lawyers reported that LLMs are useful for search, organization and discrete verifiable tasks but unreliable for final drafting or interpretation, and WarmWash and freejazz worried about billing theater and gatekeeping. Jumploops and aaimnr warned of an arms race in "slop-law" and AI-generated litigation, while sva_ and rayiner saw potential to democratize access to rights and legal research. Others framed the change as existential - either productivity for clients and firms or further concentration of wealth and erosion of professional accountability.