I'm concerned about bias in this in form of the Hawthorne effect for one group and something akin to Learned Helplessness (or self-fulfilling prophecy) for the other.
https://en.wikipedia.org/wiki/Hawthorne_effect
The Hawthorne effect can apply if they tell the group receiving the money about the expected benefits of basic income. They know the spotlight is on them. They're also getting special treatment. They might work more and do better in this situation than if basic income was the norm similar to how welfare is the norm but doesn't get people out of poverty. Whereas, it's the opposite if the control group is told about the scheme: people in an economic rut with likely negative outlook that just got slighted on basic income and will be watched to see if they will fail. They'll likely fail either by default or barely trying although some might be motivated to work harder to stick it to observers.
If Y Combinator is reading this, I suggest countering it with careful presentation to each party. Especially to the control group. Tell them they're simply being observed as part of a study on poverty in Oakland and where people end up over time. Absolutely no talk of basic income or another group.
For benefiting group, countering Hawthorne effect is less my expertise. Old trick of making both groups feel like they're getting special treatment or advantage can't apply here. Perhaps water down the effect by describing it as one of a series of experiments... with descriptions of various approaches... to help lift people out of poverty along with theory of each. Mention positive, neutral, and negative outcomes that are common on existing welfare. Mention the potential the new approach offers. Mention that they're selected to try it for a year with the results entirely on them but observed to inform future studies. Effect is still there but with a balanced presentation instead of overly-positive or specific push from researchers presenting it. Information overload might even drown out Hawthorne effect to make them focus on whatever is last thing said.
The control group I'm pretty sure on but other possibilities are speculation. Consider how to handle benefit group carefully. Regardless of methodology, I'm glad to see YC doing it as it might help some people regardless.