Make a $1T company. Make no mistakes.

The more time I spend on production AI, the more I think eval is harder than training. You can always throw more RL at a problem. The hard part is knowing what to optimize for—and trusting that your eval actually captures it.

CRB now tracks deep review agents - agents that take the time they need to perform a thorough review. Does more compute actually mean better reviews? Now we can measure it.

You may have a favorite code review tool, but is it going to survive what's to come?

Who is making the code review tool to maximize fight index? I'll start: Shai-Reviewed

Exciting work on how to control multiple aspects of LLM behavior from a model using steering. Congrats Narmeen Oozeer !

Code review is stratifying. One strata has humans still in the loop. You'd be surprised how big the other strata is.