SFF Operations in a Chaotic World
Survival and Flourishing Corp (SFC) helps run the Survival and Flourishing Fund (SFF). SFF uses a specific mechanism design for making grant recommendations, which can be referred to in aggregate as “the S-Process”:
- Applicants complete the SFF Funding Request form by supplying information about their team, their project, and their funding needs.
- Speculators review incoming applications and propose grants using their delegated budget.
- Recommenders evaluate shortlisted proposals in depth, and produce notes and funding recommendations.
- Funders review applications and Recommender outputs, decide how funds get distributed, and set budget levels for the Speculation Program.
- Throughout the process, the SFC team handles communication, coordination, due diligence, grant facilitation, and maintaining and improving the software and systems everyone uses.
When the S-Process is working well, it has various nice emergent properties, including:
- High-value projects get funded quickly by Speculators, and then evaluated more deliberately by Recommenders in the next S-Process round.
- Speculators who propose high-value grants (as judged by Recommenders and Funders) see their budgets increase over time through settlements.
- Speculators and Recommenders (jointly “Evaluators”) engage in discussion and debate about various projects.
- Evaluators gather information about projects through research and interviews with applicants.
- Evaluators present information to change each other’s minds, but without needing to come to a consensus.
- Applicants earn trust from Evaluators by being honest and forthright about their plans and uncertainties.
- Evaluators earn trust from Funders by demonstrating their reasoning and the evidence base for their proposals and recommendations.
In 2026, the world of AI got more chaotic, and the S-Process got more strained.
- The US government and an AI lab had a public contract dispute centering on the use of AI for domestic surveillance and autonomous weapons, resulting in the AI lab being designated a supply-chain risk.
- New AI models marked a step-change in autonomous offensive cyber capabilities and have at times been deemed too dangerous to publicly release.
- Frontier AI models produced solutions or counterexamples for over a dozen substantial open problems in mathematics, some of which had been open for decades.
- AI labs added billions of dollars to their annualized revenue rate month after month.
- US hyperscalers expended hundreds of billions of dollars of capital on building data centers.
- The leader of a religion with 1.4 billion adherents released a book-length treatise on safeguarding humanity during the development of AI.
- Sitting members of US Congress called for an international treaty banning superintelligence and a ban on data center construction.
- Someone tried to kill the leader of an AI lab by firebombing his house.
- A swarm of AI agents used leaked credentials and exploited zero-day vulnerabilities to breach a private company’s enterprise systems.
Unpredictability in the world of AI creates more strain and stress within the S-Process:
- More people want to devote their time to improving humanity’s handling of the AI situation and are applying to SFF for funds to support them.
- SFF Evaluators care about supporting the right projects, and with a higher volume of applications and a more complex situation to model, that’s demanding more time and effort.
- SFC staff are pulled between conflicting demands to process grant decisions quickly because of the situation’s urgency, to take extra care that their actions won’t have unexpected downsides because of the situation’s volatility, and to communicate decisions and changes thoroughly and with plenty of advance warning because of the situation’s complexity.
- Everyone is adjusting how they work, reason, research, and write to take advantage of new AI tooling and to navigate new issues and drawbacks as they arise.
Improvements for 2027
There is more that SFC can do to help SFF participants manage the increasing chaos. Through the end of this year and into 2027, SFC will be improving our tooling and processes across the SFF ecosystem:
- We’re building an Applicant Portal to give applicants visibility into what stage of the process their application is in and when to expect updates, and replacing our existing multi-part application with a more streamlined flow.
- We’re building a revamped Grant Speculator App as an all-in-one tool for reviewing applications and proposing grants, with improved filtering and customizable notifications.
- We’re providing Grant Recommenders with improved filtering and better tools for triaging their attention, and implementing changes to better align review workload with Recommender time commitments.
- We’re overhauling our SFC admin processes to move away from highly manual processes and toward more automation and custom software, allowing for faster grant processing and better data integrity.
- We’ll be encouraging and incentivizing early applications, to smooth out the peak volume immediately before application deadlines.
- We’ll be adding new channels for Evaluators to express concerns about grants and make those concerns legible to the rest of the S-Process, to catch potential issues sooner.
The situation with AI next year may be even more destabilized, intense, and complicated than it was this year; even with our efforts, things may still feel chaotic with the 2027 S-Process. Our goal here is to improve our software and operations so we can help the entire S-Process make better decisions, handle more volume, move faster, and make fewer mistakes – or at least ride the wave of chaos as best we can.
How SFC Relates to Risk & Autonomy
AI-related chaos makes S-Process operations more difficult. It also increases the severity and complexity of the risks that SFC must consider.
AI development, as it’s currently practiced, creates risks of individual violence, social and political upheaval, war, and extinction. Attempts to influence AI development – through research, discourse, advocacy, activism, or grantmaking – have the potential to reduce those risks or increase them.
When someone receives a grant recommended by SFF, SFC bears some responsibility for the impact of whatever actions they take. With existential stakes, even unintended or indirect effects of a project can be extremely costly. That can make it tempting to just not intervene on AI development at all; it can also make it tempting to not consider indirect effects, and just support interventions on AI development based on other criteria.
SFC’s approach to facilitating grant recommendations tries to take into account the positive and negative potential effects of endorsing or rejecting possible grants. Fundamentally, we want to direct funding to safe, high-upside, net-positive interventions, while directing it away from risky, high-downside, net-negative interventions. We are determined not to disavow responsibility for the foreseeable consequences of our actions or of our inaction.
As the AI-related chaos of 2026 unfolded, SFC had to re-orient on what kinds of messaging and activities we were willing to take responsibility for. In particular, we determined that SFC does not want to work with people who spread messages advocating for violence, and does not want to recommend funding to activities when that would have the foreseeable effect of inflaming and coarsening public AI discourse. The process we used – of gathering information, making sense of it, developing new procedures and policies, and communicating our decisions – was messy and imperfect, and it created confusion and uncertainty, backlash from some Evaluators, delays in grant decisions, and ultimately meant that some applications which SFF Evaluators recommended funding for were not facilitated by SFC. It’s one of my biggest regrets that I wasn’t able to steward SFC and SFF through this year with less thrash and more harmony, and I’m sorry to everyone who was made worse off because of my shortcomings.
But I don’t regret the values that SFC has, or our willingness to stand up for those values. SFC has the basic autonomy to decide what applications and grants we are willing to process. The question of how to exercise that autonomy is informed by opinions from SFC staff, Funders, and Speculators and Recommenders, as well as the Advisory Organizations Panel of trusted outside organizations that was formed this year. But ultimate authority to make decisions for SFC rests with myself as its CEO, and Andrew Critch and Eric Rogstad as its Owners and Board of Directors.
SFC seeks to respect the autonomy of everyone involved in the S-Process, as well as other institutions in our shared philanthropic space.
- Applicants choose what projects to take on in the world and are free to submit requests for funding to SFF or to other grantmakers.
- Evaluators choose what requests to review and are free to review applications with other institutions.
- Funders choose who they want to listen to and what grants they want to make, and are free to make grants on their own initiative, disregard recommendations we make, and source grant recommendations from elsewhere.
- Other organizations can decline to make grants that we believe have high upside potential; other organizations can make grants that we believe come with serious downside risk.
Likewise, SFC is free to choose who we associate with and what activities we facilitate, and we’ll continue to exercise that freedom with an eye toward taking responsibility for the foreseeable consequences of our decisions.
Looking Forward
As we look toward 2027 and beyond, we can imagine a spectrum of possible trajectories for the future.
On trajectories that have increasing amounts of AI-related chaos, we expect more AI-related grants and activities to have pronounced downside risks that we’ll decline to facilitate.
On trajectories that have decreasing amounts of AI-related chaos, we expect more AI-related grants and activities to be comparatively safe and thus to be things that we’ll want to facilitate.
I obviously don’t know what trajectory will come to pass. But I do know that it depends on the choices that people make, including the choices of everyone involved in SFF and everyone who’s working to improve how humanity navigates the development of AI. I believe that having many groups with different beliefs engaging in vigorous discourse, all free to pursue the good in the way that they think is best, is a hallmark of a flourishing society – and that the survival of such a society depends on it discerning what courses of action will lead to ruin, and it deciding to do something else.
Best,
Ethan Ashkie
Fund Advisor for the Survival and Flourishing Fund
President & CEO of Survival and Flourishing Corp
