Haha indeed!

On a more serious note, just as swyx mentioned in a comment further up, we do believe a lot of the challenges of reliably operationalising agents boil down to data. All of which is non-obvious to AI engineers (besides Frontier Labs gathering/generating data for model training).

What the right shape is - we are all figuring it out. Happy to trade notes.