I'd like to see a world where AI agent setups are managed per user/employee instead of administered through an employer. So many problems to solve around data privacy and access and it will never happen.
@investing_law No one wants to hear it but firms need to start handing out ownership positions much earlier in career like tech companies. It’s the best way to get incentive alignment, foster connections, and develop a bizdez toolset early in career.
Yeah that makes sense. I’ve been spending time thinking through an oss version of CMA for many of these same reasons and was disappointed in the lack of offerings but it’s new (although there are some promising new projects like https://t.co/eGdzBbb7Eh)
Would love to read more about the engineering behind it when/if you guys decide to share.
@gabepereyra Yeah that aligns better with what I expected on the ZDR point. Some 3rd party somewhere has to store the data to do agentic workflows unless you do on prem but managing that risk with one vendor (eg, Harvey) is easier than a vendor and all of the managed platforms they rely on.
Agreed, I expect the gap to exist I just also expect it to narrow over time but not sure it will ever be closed.
I think the point on compute is a big one too. Dario said not too long ago that os ai models are not really os bc if he gave you the weights for a large model you wouldn’t have the resources to run it.