SFF’s Speculation Grants Program

Last updated: February 28th, 2025

Expedited grants outside of S-Process

SFF Speculation Grants are expedited grants organized by SFF outside of our grant-recommendation process (the S-Process). “Speculation Grantors” are individuals with budgets to make these grants. Each Speculation Grantor’s budget grows or increases with the settlement of budget adjustments that we call “impact futures” (explained further below). Currently, we have a total of ~35 Speculation Grantors, with a combined budget of approximately $16MM (up from $4MM initially). Our process and software infrastructure for funding these grants were co-designed by Andrew Critch and Oliver Habryka.

In order for an applicant to be guaranteed eligibility for consideration in an S-Process Grant Round, they must be awarded a Speculation Grant.

Here’s how it all works, in more detail:

How to request a Speculation Grant (application instructions)

To request a Speculation Grant, complete and submit the SFF Funding Rolling Application. Filling out this application submits a Speculation Grant Request for expedited funding as well as a request for consideration in the next S-Process Grant Round.

Submitting by the S-Process Grant Round deadline is required for Speculation Grantors to settle a Speculation Grant to you in the upcoming round, so your chances of receiving a Speculation Grant are improved by submitting early. See the SFF announcements page for grant round announcements and application information, including the deadline. You can still apply even if there is no currently open grant round.

How quickly are Speculation Grant requests evaluated?

Each Speculation Grantor is individually free to reach out to you and make a Speculation Grant (though they are expected to notify the other Speculation Grantors of their decisions), and in principle such a grant could be made as quickly as one week after the grant request. However, there is no fixed timeframe within which a decision will be made.

But what is a Speculation Grant?

Speculation Grants are expedited grants approved by a person called a “Speculation Grantor”. Each Speculation Grantor has a limited budget for approving Speculation Grants (typically $400k). Each grant is evaluated for quality and entry in a subsequent S-Process Grant Round, and the Speculation Grantor’s budget is then replenished by an amount depending on the quality of the grant.

Speculation Grants can be intended for grants that are somewhat time sensitive and cannot wait for a few months until the next SFF round distributes grants. Decisions may be made

• as soon as one week after the Speculation Grant request is made,
• as late as 14 days after the Grant Round application deadline (if a Speculation Grantor decides to wait to see what other applicants to the round are available to receive Speculation Grants, and then chooses your application), or
• possibly even later (if there’s a logistical delay with approving your grant).

Currently, all our Speculation Grants are funded by donations from Jaan Tallinn, and can be intended for funding requests that can’t wait to be processed by the most recently announced S-Process Grant Round.

To get money faster, you have to provide more information, not less. In particular, you still have to submit a full application, to be evaluated by the S-Process Grant Round after the fact of the grant. In general, if you receive a Speculation Grant for $Y from SFF, and the close of SFF’s next S-Process Grant Round recommends that your organization receive a total of $Z from that round, then the $Y is counted against the total $Z recommendation. I.e, if the amount distributed to you in Speculation Grants before the close of the round is (was) $Y, then we’ll recommend an additional max(0, $Z - $Y) to be distributed to you after the close of the round.

From a myopic perspective, the “max(0, –)” formula above means that a Speculation Grant guarantees you receive at least $Y in total, which is nice, although if the grant decision turns out to be evaluated poorly by the S-Process Grant Round (see below), it can hurt your reputation and the reputation of the Speculation Grantor who recommended it. On the other hand, the Speculation Grant can also provide a positive boost to your reputation if it’s evaluated positively, because it will lead to an increase in budget for the Speculation Grantor who recommended the grant, which is a happy and exciting thing to see happening.

Who’s involved?

Currently, ~35 individuals are serving as Speculation Grantors, with an average budget of around $400,000 each from which to approve Speculation Grants for SFF to make. The following Speculation Grantors have agreed to share their names and/or contact information publicly:

• Andrew Critch (no public email listing)
• Benjamin Todd <ben@80000hours.org>
• David ‘davidad’ Dalrymple (no public email listing)
• David Kristoffersson (no public email listing)
• David Krueger (no public email listing)
• Deger Turan (no public email listing)
• Eliezer Yudkowsky <yudkowsky@gmail.com>
• Irene Solaiman (no public email listing)
• Jessica Taylor (no public email listing)
• Justin Shovelain (no public email listing)
• Nell Watson (no public email listing)
• Oliver Habryka <oliver.habryka@gmail.com>
• Roman Yampolskiy (no public email listing)
• Samuel Hammond (no public email listing)
• Seth Lazar (no public email listing)
• Steve Omohundro (no public email listing)
• Tsvi Benson-Tilsen <tsvibtcontact@gmail.com>
• Zvi Mowshowitz <thezvi@gmail.com>

Each Speculation Grantor’s budget increases or decreases over time based on the value of their grants as evaluated by the settlement of “impact futures” by the S-Process, described further below. Our Speculation Grants are primarily administered through SFF’s DAF, which has been funded via donations from Jaan Tallinn. Some Speculation Grants, for instance grants to for-profit organizations, administered through Survival and Flourishing Corp and also funded by Jaan Tallinn. In the future, additional S-Process Funders might join.

How do SFF’s impact futures work, logistically?

SFF tracks the Speculation Grants we make in the form of (non-binding) “impact futures” that “settle” in the form of budget adjustments for our Speculation Grantors. Here’s how that works internally, using an example.

(All of this describes how we expect things to work “by default”, but like all of SFF’s announcements, none of this description should be taken as binding, and might change considerably if we notice problems with the settlement process.)

Suppose Alice is an SFF Speculation Grantor with a starting budget of $200k. That “budget” is just an entry in a spreadsheet managed by SFF, which doesn’t really belong to Alice; it represents how much we’re planning to listen to Alice’s (non-binding) suggestions to us. In other words, Alice never receives any money; just an invitation to recommend grants for SFF to make from the budget.

Next, suppose Alice sees a request for a Speculation Grant from some organization called the Beta Institute. Alice decides to recommend a $50k Speculation Grant from SFF to the Beta Institute, and SFF makes the grant. Now, three things are recorded:

• $50k is subtracted from Alice’s $200k Speculation Grant budget, so she can recommend another $150k of Speculation Grants from SFF before running out.
• A grant deduction of $50k to the Beta Institute is recorded in our recommendation process for Jaan Tallinn (as he is currently the sole donor supporting our Speculation Grants program). This means that the next time SFF’s S-Process produces a recommendation for how much Jaan should give to the Beta Institute, $50k will be subtracted from our (non-binding) recommendation.
• An impact future for granting $50k to the Beta Institute is recorded for Alice’s grant-making (recommendation) budget. Similar to a stock future, the impact future will settle to (i.e., adjust Alice’s budget by) a value settled by a future evaluation. That value is the product of two factors: the utility value of the grant as evaluated by Jaan’s delegates in the next S-Process Grant Round, and a normalizing factor (Jaan’s total budget for settling impact certificates). See below for the details of how these factors are used.

Next, suppose the next S-Process Grant Round recommends that Jaan should grant a total of G dollars to the Beta Institute. By default, Jaan’s grant deduction and Alice’s impact future are “settled” by adjusting Alice’s budget (no money actually flows to Alice the person) as follows:

1. Grant deduction settlement. Jaan makes a grant to the Beta Institute of size max(0,G-$50k), and the $50k grant deduction is marked “settled”. In particular, it does not carry forward to future rounds.
2. Impact future settlement. Alice’s impact future “settles” to a value determined by a special run of the S-Process simulator in “evaluation mode” , where Jaan’s S-Process grant Recommenders enter dollar values for the goodness of each dollar granted in the round, and their opinions are combined using marginal value functions and a total impact future budget specified by Jaan. (If Alice is also serving as a grant Recommender for the S-Process itself, she is “recused” in that her values are purged from the system during the evaluation of her future certificate). The settlement value of the impact future and Alice’s resulting budget are then calculated as follows:
   • For a given future certificate cert, we first calculate its value in “first dollars”, i.e., relative to the value of the first dollar granted in the round:
     • fd_value(cert) = the total dollar value of the first 50k Jaan-dollars recommended for the Beta Institute, as calculated by the S-Process simulator with Alice recused (if she’s a Recommender), other Funders removed, and no discount factor or “hold funding” option for Jaan. The fd_value of a given certificate cert will be some number less than 50k, because the most valuable Jaan-funded dollar granted in the round is defined to be valued at $1 in the simulator (first-dollar values are renormalized and capped at 1). The value of the certificate will be proportional to this number.
   • Next, we add these up to get the total “first dollar value” of all the impact certificates:
     • total_fd_value := sum[fd_value(cert)) for cert in all_certs]
   • Next, Jaan, chooses a total budget for settling the impact certificates in the given round.
     • settlement_budget = Jaan’s chosen total budget for settling (like “buying back”) impact certificates. If this is greater than the total amount granted through the Speculation Grants under evaluation, the Speculation Grant program grows in total size; otherwise it shrinks. (Jaan also has the option of delegating this number to the S-Process itself to choose.)
   • Next, we choose a settlement normalization factor to ensure the total valuation of certificates adds up to the settlement budget:
     • settlement_normalizer = settlement_budget / total_fd_value
   • Finally, the settlement value of each certificate is adjusted up or down to make the total budget match Jaan’s chosen budget for them, and Alice’s budget is adjusted by the settlement value:
     • settlement_value(cert) := fd_value(cert)*settlement_normalizer
     • Alice_budget = $150k + settlement_value(cert)

In summary, if Jaan’s and his advisors’ marginal value functions together say that the $50k grant was worth, say, 33k “first Jaan dollars”, and Jaan chooses a total settlement budget equal to 3x the total value of the Speculation Grants measured in “first Jaan dollars”, then Alice’s certificate settles to 3.0*$33k = $99k, and her new budget is $200k - $50k + $99k = $249k.

So overall, to the extent that Alice wants to grow her Speculation Grant budget, her best strategy is to make grants that the next S-Process Grant Round will consider to be highly valuable. If she wants, she can also use some or all of her budget granting to things that the S-Process is unlikely to value, and it’s up to Alice to decide how and what to prioritize.