My setup allows bug fixes and improvements where input > output is known quantities to ship anytime. However for anything that has a subjective quality bar I still want to user test it manually to ensure that features I ship to users feels good by my subjective discernment.
If something breaks, you likely want to be able to at least react and roll-back?
The agents can check this and mitigate.
I ship overnight all the time now. I have a huge test set that runs before we ship. Users find updates each morning which we think feels nice.
because responsible orgs shipping real software don't do the eyes free coding (different from hands free)
What is the difference between hands free and eyes free? Could you explain a bit more?
eyes-free is ralph loop, blindly trusting Ai output, swarms making huge swaths of changes
hands-free is not typing, but staying in the loop, evaluating claims, and taking time to look through the code without an agent