Arguments along these lines are giving me quite a headache aswell. That being said I would like to push back against this line of reasoning a bit:

• According to the same argument, the hypothesis "only I am sentient" beats the hypothesis "all human beings are sentient" approximately 7 billion to 1 in explaining the observed evidence (i.e. the evidence that you are born as yourself). Assuming that "all humans" includes all human beings throughout humanity's history this ration plummets even further. (I am not actually trying to argue that you are the only sentient creature since there are good reasons to assume that other human beings are sentient aswell (for example the similarities in brain structure) but that this line of logic is to be considered with caution.)
• Only looking at a single sample is not a consistent estimator. The definition of a consistent estimator doesn't map super nicely onto our current setting since it requires the number of samples to go to infinity. As far as I am aware there is no way to be born multiple times (or for that matter an infinite number of times) to gather more data. Wikipedia has this example where only the last value of a measurement series is taken to estimate the expected value of a random variable. In my opinion this maps fairly nicely to our setting, where we can only access one data point to begin with. As shown (see link), the extimation process doesn't converge (as it is not a consistent estimator) and we should therefore expect that it doesn't give us any reliable insight into the underlying random variable. 

Thank you for sharing!

(Simply giving this link was really valuable already. I will put my main takeaway points from this post here, but I do not expect a response in case you do not want to participate in this discussion again as it appears you already did so 3 years ago :'D.)

• To me, it appears as if the authors mainly agree that diversification (within one cause area) is motivated by an attempt to maximize utility, though they disagree on the degree to which diminishing returns of investment (and therefore the role of diversification) actually matter in practice. I briefly want to point out that even though this idea is obvious from an EA perspective the literature on donor coordination problems does a very poor job in capturing this intent.
• I agree that donation funds help reduce coordination problems. However, assuming there is more than one grantmaker, this just shifts the "burden of aggregating different opinions" from the general public into the organization since grantmakers still need some mechanism to reconcile their differing beliefs. That said, I don't know enough about typical grantmaking processes to judge whether grantmakers differ significantly enough in their individual assessments for this to matter in practice.

I agree with the Overall statement of this Post. Regarding a "GiveWell of X" type of organization I believe it would have to function quite differently, ideally only working on-demand instead of doing broadly aimed research for the following 2 reasons:

•  The target group of such an organization would have to be succeptible enough to cost-effectiveness arguments to be moved by their research, but resilient enough not to already pivot to a GiveWell charity instead. That sounds very nieche. Starting out with the mission of an established organization and convincing them that there is a better way to achieve their goal sounds like a more promising approach than "more broadly aimed research".
• In case this organization presents itself similar to GiveWell (easily browsable website with their main results, etc.) this could dillute the perceived effectiveness discrepancy to GiveWell charities. But then again most donors with the tiniest bit of alignment to EA probably Look at the numbers too much for this to become an issue.

I love this concept and I really hope it gets the momentum required to really take off!

One minor bug report: I was trying to sign up to GoodWallet in Firefox via my Google account, which caused the website to already fill out [first_name][second_name] as my suggested GoodWallet-pagename. However, this string contained more than 20 characters, which made it impossible to create this wallet. Trying to delete this string to manually choose a smaller pagename did not work, which got me stuck in this step of signing up. Retrying the same process in Chrome worked just fine. (Let me know if you need more details to recreate the bug/if there is a dedicated way to report these sorts of issues properly.)

As already suggested by many other users, I also believe that increasing the tax efficiency is crucial. I am no expert what the best way of achieving this would be. However, I feel like my own likelihood of using this tool would drastically increase with its (perceived) monetary efficiency.