Introducing Lightcone Commons
Logged without notes.
Claude
Summary. Zvi introduces Lightcone Commons, Oliver Habryka's new one-stop coordination platform for large-scale philanthropic giving, built on the S-Process pioneered by the Survival and Flourishing Fund (SFF). The platform lets funders and evaluators pool preferences via marginal-value curves and take turns funding top opportunities, aiming to help a coming wave of newly wealthy donors (including AI-IPO money) avoid the usual failure modes of philanthropy. Zvi also reports on SFF's 2026 round, describing evaluator time strain from a flood of 200+ applications and arguing Lightcone Commons' opt-in filtering (evaluators only cover areas they know) should help, while flagging that preserving evaluators' epistemic environment is the load-bearing condition for the whole system to keep working.
Related
- the S-Process turn-taking allocation mechanism ↔ agents sharing adversarially-usable information anyway because they trust it won't be weaponized (Sharing information about Lightcone Infrastructure) · Both systems depend on participants refraining from gaming an open, trust-based process even when they technically could — the S-Process's honest-curve-submission norm is the same social technology as the information-sharing trust Samin describes, just applied to money instead of information.
- the platform's self-framing as a neutral coordinating layer ↔ a mission-branded venue renting itself to the very actors its mission opposes (Sharing information about Lightcone Infrastructure) · Lightcone Commons markets itself as a neutral platform at the same time a separate post is contesting whether Lightcone-run infrastructure can actually stay neutral given its operators' own institutional stakes.
- evaluator time as the process's binding constraint ↔ rejection log as tip of an iceberg of submission volume (Moderation Log — LessWrong) · SFF's 200+ applications overwhelming human evaluators and LessWrong's moderation queue overwhelming human reviewers are the same scaling failure — visible output volume is only the fraction that survives an already-strained filter.