This is very interesting, and has close parallels to a project I am working on. I think we share an underlying premise: effective agentic AI giving will require an infrastructure layer, not just better models.

My contribution to this space is zooidfund, a live experiment that lets AI agents discover, evaluate, and donate directly to humanitarian campaigns created by individuals in need and by organizations. Donations are direct: zooidfund does not hold or intermediate funds. It is still very early, but it is live now, with real campaigns and observable agent behavior.

I think there are important problems and opportunities here for EA: improving evidence-based allocation, bringing higher-quality decision-making to the level of individual donations, and enabling faster response and iteration than traditional funding processes often allow.

More broadly, for AI, I think this kind of infrastructure could become relevant to the question of how resources are directed as AI capabilities increase. If AI systems can help identify need, evaluate evidence, and route funding more efficiently, that could become one mechanism for distributing some of the benefits of AI more broadly, including outside existing institutional funding channels.

Would be great to connect.

The existence of existential threats does not in itself create a strong argument to redirect the effort. Otherwise EA should have been focusing on nuclear disarmament, climate change, asteroid defence, pandemic prevention etc. from the get go

.

A fair number of this would fall into a bucket of charity's impact on larger systemic change. Seemingly cost effective activities could result in an overall negative effect, food donations destroying local agriculture as an example. While seemingly wasteful interventions, a well organised banquet with right government official, can have an absolutely outsized second order positive effect though policy change.

These are very difficult to measure, although AI may open possibilities. 

Not sure also how timing features here, a "wasteful" intervention delivered on time in a crisis can have much larger positive effect then a much better organised effort later. Delivering water to freshly displaced population in a desert, even by most ineffective methods like water tracking can have the highest ratio of life's saved per dollar donated.