# Holden Karnofsky Holden Karnofsky (b. 1981) is the **co-founder of GiveWell (2007) and founding executive of Open Philanthropy (2014-2021)**, and one of the two or three most structurally important operational figures in the Effective Altruism movement's institutional build-out across the 2007-2024 period. A Harvard philosophy graduate (2003) who began his career at the [[Bridgewater Associates]] hedge fund, Karnofsky left finance in 2007 to co-found GiveWell with colleague Elie Hassenfeld — initially as an evening-and-weekends effort to evaluate charities with the rigor Wall Street applied to investments, then expanding into the principal evidence-based charity evaluator the EA movement built around. In 2014, under the "GiveWell Labs" branding that became Open Philanthropy, Karnofsky structured the partnership with [[Dustin Moskovitz]] and [[Cari Tuna]] that turned Moskovitz's Facebook-derived fortune into the largest institutional EA grantmaker — eventually scaling to ~$4 billion in cumulative grants under Karnofsky's presidency. Karnofsky stepped down from the Open Philanthropy CEO role in 2021 (Alexander Berger took over) and from 2022 onward has focused on AI safety work, including a board seat at [[Anthropic]] where his wife [[Daniela Amodei]] is President and brother-in-law [[Dario Amodei]] is CEO. Karnofsky matters for the KB for three reasons. First, he is the **principal operational architect of modern EA philanthropy** — the figure who converted the MacAskill-Ord intellectual framework into the institutional grantmaking apparatus that actually directed billions of dollars across EA's 2014-2024 period. More than any other single figure, Karnofsky is the person who made EA operationally functional at scale rather than just intellectually significant. Second, his **Cold Takes blog** (launched 2021) and particularly its ***Most Important Century*** essay series is one of the most-cited expositions of the [[Longtermism]] argument for AI safety as the defining moral priority of the present moment — translating Bostrom, MacAskill, and Ord's academic framings into a readable long-form blog format that has substantially shaped the EA-to-AI-safety pipeline. Third, his **Anthropic board seat and family connections to Anthropic's leadership** (Daniela Amodei is his wife; Dario Amodei is his brother-in-law) make him a documented case study in how the EA philanthropic infrastructure and the AI safety commercial-AI-lab infrastructure have become structurally intertwined — the boundary between "EA funders" and "AI safety commercial entities" has substantially blurred through exactly these kinds of family and institutional linkages. ## Background - Born 1981; raised in suburban Philadelphia - BA from Harvard University (2003, social studies) - Analyst at Bridgewater Associates (2003-2007) under [[Ray Dalio]] - Co-founded GiveWell with Elie Hassenfeld in mid-2007, initially as a side project during their Bridgewater tenure; left Bridgewater full-time shortly thereafter - Executive Director, GiveWell (2007-2014) - Executive Director, then CEO, Open Philanthropy (2014-2021) - Director of AI Strategy, Open Philanthropy (2022-2024) - Anthropic Board member (2024-present) - Married to Daniela Amodei, President of Anthropic; brother-in-law to Dario Amodei, CEO of Anthropic - Based in San Francisco ## GiveWell Founding (2007) GiveWell emerged from Karnofsky and Hassenfeld's dissatisfaction with the absence of rigorous charity evaluation. As Bridgewater analysts accustomed to evaluating hedge fund managers on transparent quantitative criteria, they were struck by how opaque the charitable sector was — charities were evaluated on overhead ratios and Charity Navigator-style financial metrics rather than on whether their interventions actually worked. GiveWell's methodology, developed 2007-2014, became foundational for EA: - **Intervention-level focus**: Evaluate charities based on the underlying intervention (bed nets, deworming, direct cash transfers) rather than on organizational characteristics - **Cost-effectiveness modeling**: Quantitative estimates of dollars-per-life-saved or dollars-per-DALY-averted across interventions - **Top Charities list**: An annually updated short list of charities the organization judges to be most cost-effective, which directed hundreds of millions of dollars across the 2010s - **Radical transparency**: Publication of full reasoning, including criticisms of recommended charities and the organization's own errors The GiveWell methodology positioned the organization at the center of the emerging EA movement — when [[William MacAskill]] and [[Toby Ord]] founded [[Centre for Effective Altruism]] in 2011-2012, GiveWell was already the operationally-proven evidence-based-giving organization the intellectual movement could point to. ## Open Philanthropy Launch (2014) In 2014, GiveWell launched "GiveWell Labs" — a project to expand beyond global health into additional cause areas. The Labs project rapidly became the principal vehicle for Dustin Moskovitz and Cari Tuna's philanthropic giving, and was formally separated as Open Philanthropy Project (later Open Philanthropy) in 2017. Under Karnofsky's leadership (2014-2021), Open Philanthropy expanded to cover: - **Global health and development**: Continuing the GiveWell framework at scale - **Farm animal welfare**: Substantial grantmaking to cage-free egg campaigns, corporate animal welfare commitments, and alternative protein research - **Scientific research**: Biosecurity, pandemic preparedness, and specific scientific priorities - **Criminal justice reform**: U.S. domestic grants around incarceration reduction - **Potential risks from advanced AI**: The longtermist core; grants to MIRI, OpenAI (the famous 2017 $30M grant), later Anthropic, and AI safety research - **Global catastrophic risks**: Biosecurity, nuclear security, great power conflict By 2021 (when Karnofsky stepped down as CEO) Open Philanthropy had deployed ~$1 billion in cumulative grants; the total has continued to grow to ~$4 billion by 2024. Alexander Berger succeeded Karnofsky as CEO in 2021, with Karnofsky remaining as Director of AI Strategy. ## *Most Important Century* and Cold Takes (2021-) Karnofsky launched his personal blog *Cold Takes* in 2021 (after reducing his Open Philanthropy operational load). Its signature essay series, *The Most Important Century*, published in 2021-2022, has been one of the most widely-circulated expositions of the longtermist AI-risk argument outside academic publication. The core argument: - **Transformative AI could plausibly arrive this century**, driven by the continuation of compute-scaling trends and ongoing AI capability gains - **Transformative AI would produce economic growth rates comparable to the Industrial Revolution** compressed into decades rather than centuries - **The transition period is uniquely high-stakes**: The values embedded in the first transformative AI systems may shape the entire long-term future - **Therefore, AI alignment and AI governance are plausibly the most important work of the present generation** The essays translated [[Nick Bostrom]]'s *[[Superintelligence (Book)]]* (2014) and other academic longtermist work into blog-length argument suitable for a broader audience, and have been influential in both the EA-to-AI-safety pipeline and the broader "AI is the defining issue of our time" discourse that has become dominant in Silicon Valley since 2022. ## Anthropic Transition (2022-) Karnofsky's post-2021 trajectory has been toward AI safety and specifically toward Anthropic: - **Anthropic funding**: Open Philanthropy's late-2010s grants to OpenAI, and subsequently to Anthropic after its 2021 founding, were substantially driven by Karnofsky's AI strategy work - **Family connections**: Karnofsky's wife Daniela Amodei is Anthropic's President; her brother Dario Amodei is CEO. The family structure — two siblings leading a major AI company while married to the principal AI-safety philanthropic strategist — is documented in public reporting and is a case study in how EA philanthropic infrastructure and AI commercial entities have become structurally intertwined - **Board seat (2024)**: Karnofsky joined the Anthropic board in 2024, formalizing the relationship - **Ongoing AI safety writing**: Cold Takes continues publication, though at reduced frequency ## Connection to the SBF / FTX Period Karnofsky's relationship to the [[Sam Bankman-Fried]] / FTX collapse is a documented piece of his record: - Open Philanthropy and FTX Future Fund were the two principal institutional funders of EA-aligned longtermist work in 2021-2022 - After FTX's November 2022 collapse, Karnofsky and Open Philanthropy absorbed or backstopped some grants that had been paid or committed by FTX - Karnofsky has publicly stated that he had some interactions with SBF but was not centrally involved in FTX governance, and has been critical of the post-FTX governance failures at CEA and EA-adjacent organizations ## Connection to the KB Ecosystem - **[[Effective Altruism]]**: Principal operational figure in the EA movement's institutional build-out. - **[[Open Philanthropy]]**: Founder and executive through 2021; remains one of the principal voices in its current strategy. - **[[GiveWell]]**: Co-founder; the operational predecessor of Open Philanthropy. - **[[Dustin Moskovitz]] / [[Cari Tuna]]**: Long-term philanthropic partners; the Moskovitz fortune is substantially channeled through Karnofsky's institutional apparatus. - **[[Longtermism]] / [[TESCREAL]]**: Principal blog-and-grantmaking vehicle for translating longtermist intellectual work into institutional action. - **[[Anthropic]]**: Board member; spouse is President and brother-in-law is CEO. - **[[William MacAskill]] / [[Toby Ord]]**: Intellectual counterparts; Karnofsky is the operational-infrastructure figure to their intellectual-framework roles. - **AI safety pipeline**: Principal strategist for the EA-funded AI safety research ecosystem. ## See Also - [[William MacAskill]] - [[Toby Ord]] - [[Dustin Moskovitz]] - [[Cari Tuna]] - [[Open Philanthropy]] - [[GiveWell]] - [[Centre for Effective Altruism]] - [[Effective Altruism]] - [[Longtermism]] - [[TESCREAL]] - [[Anthropic]]