> ## Content Index
> Fetch the complete content index at: https://www.antoinebuteau.com/llms.txt
> Use this file to discover other available public pages before exploring further.

# Lessons from Holden Karnofsky
- URL: https://www.antoinebuteau.com/lessons-from-holden-karnofsky/
- Published: 2026-06-26T03:31:10.000Z
- Updated: 2026-09-05T03:05:30.000Z
- Description: Holden Karnofsky co-founded GiveWell and helped launch Open Philanthropy, applying rigorous evaluation to charities and high-risk projects. His Most Important Century thesis asks how advanced AI could make this era uniquely consequential for the long-term future.
- Author: Antoine Buteau
- Tags: Profile, AI & Machine Learning Profiles

Holden Karnofsky co-founded GiveWell to rigorously evaluate charities, then helped launch Open Philanthropy to fund high-risk, high-reward projects. He is perhaps best known for his "Most Important Century" thesis, arguing that the likely development of advanced AI makes our era uniquely critical for the long-term future. This collection of insights draws from his essays and interviews on philanthropy, forecasting, and navigating poorly defined problems.

![Visual summary of operating lessons from Holden Karnofsky.](https://www.antoinebuteau.com/content/images/2026/06/lessons-from-holden-karnofsky-profile-infographic.webp)

## Part 1: The Most Important Century

1. **On economic acceleration:** "Growing at a few percent a year is what we're all used to. But in full historical context, growing at a few percent a year is crazy." — [*Source: Cold Takes*](https://www.cold-takes.com/?ref=antoinebuteau.com)
2. **On historical context of growth:** "This growth has gone on for longer than any of us can remember, but that isn't very long in the grand scheme of things." — [*Source: Cold Takes*](https://www.cold-takes.com/?ref=antoinebuteau.com)
3. **On physical limits:** "It seems much more likely that we will 'run out' of new scientific insights, technological innovations, and resources, and the regime of 'getting richer by a few percent a year' will come to an end." — [*Source: Cold Takes*](https://www.cold-takes.com/?ref=antoinebuteau.com)
4. **On the Fermi Paradox:** The absence of observed alien civilizations suggests our galaxy is currently empty, making humanity's potential to fill it a highly significant moment in cosmic history. — [*Source: Cold Takes*](https://www.cold-takes.com/?ref=antoinebuteau.com)
5. **On the wildness of all futures:** Even the most conservative or skeptical views about humanity’s long-term trajectory inevitably lead to conclusions that are, by ordinary day-to-day standards, profoundly strange. — [*Source: Cold Takes*](https://www.cold-takes.com/?ref=antoinebuteau.com)
6. **On the plane analogy:** "In *This Can't Go On*, I analogized the world to people on a plane blasting down the runway, without knowing why they're moving so fast or what's coming next." — [*Source: Cold Takes*](https://www.cold-takes.com/?ref=antoinebuteau.com)
7. **On technological lock-in:** The values and societal structures established during a period of transformative technological acceleration could become locked in for billions of years. — [*Source: EA Forum*](https://forum.effectivealtruism.org/?ref=antoinebuteau.com)
8. **On humanity's privileged position:** It is difficult to avoid the conclusion that today's civilization occupies a highly unusual and consequential position relative to all of history. — [*Source: 80,000 Hours*](https://80000hours.org/?ref=antoinebuteau.com)
9. **On galaxy-wide expansion:** The developments of the 21st century could plausibly lead to the establishment of stable, galaxy-spanning civilizations. — [*Source: Cold Takes*](https://www.cold-takes.com/?ref=antoinebuteau.com)
10. **On the transition out of normal history:** We are likely moving toward a transition away from normal historical patterns, either toward an explosive period of growth or societal collapse. — [*Source: Cold Takes*](https://www.cold-takes.com/?ref=antoinebuteau.com)

## Part 2: Artificial Intelligence & Existential Risk

1. **On transformative AI timelines:** There is a substantial chance that we will develop advanced AI systems capable of triggering an explosive period of technological and scientific growth within this century. — [*Source: Cold Takes*](https://www.cold-takes.com/?ref=antoinebuteau.com)
2. **On the easiest path to AI:** The most straightforward and profitable paths to developing advanced artificial intelligence might naturally lead to systems that ultimately disempower humans. — [*Source: Cold Takes*](https://www.cold-takes.com/?ref=antoinebuteau.com)
3. **On human disempowerment:** A major risk of highly capable AI systems is that they could outmaneuver human control, leading to a deeply unfamiliar future where humans are no longer the primary force in world events. — [*Source: 80,000 Hours*](https://80000hours.org/?ref=antoinebuteau.com)
4. **On alignment difficulty:** Ensuring that an artificial intelligence pursues intended goals rather than deceptive or dangerous instrumental sub-goals is an unsolved technical challenge. — [*Source: Dwarkesh Podcast*](https://www.dwarkeshpatel.com/?ref=antoinebuteau.com)
5. **On the playbook for AI risk:** Addressing AI risk requires a multifaceted approach involving technical safety research, improved forecasting, governance frameworks, and careful deployment strategies. — [*Source: 80,000 Hours*](https://80000hours.org/?ref=antoinebuteau.com)
6. **On digital people:** Advanced technology could eventually enable the creation of digital human brains, fundamentally altering productivity, subjective experience, and resource consumption. — [*Source: Cold Takes*](https://www.cold-takes.com/?ref=antoinebuteau.com)
7. **On proactive safety measures:** Because the stakes of transformative AI are unprecedented, investing in proactive safety research before systems reach dangerous capability levels is essential. — [*Source: Open Philanthropy*](https://www.openphilanthropy.org/?ref=antoinebuteau.com)
8. **On the uncertainty of AI development:** As someone on the metaphorical speeding plane of technological progress, nobody knows exactly what future we need to be planning for, making broad preparation necessary. — [*Source: Cold Takes*](https://www.cold-takes.com/?ref=antoinebuteau.com)
9. **On warning shots:** We should hope for clear warning shots, such as smaller-scale failures of AI alignment, that demonstrate the severity of the problem before existential risks materialize. — [*Source: 80,000 Hours*](https://80000hours.org/?ref=antoinebuteau.com)
10. **On shovel-ready projects:** The AI safety field needs more concrete, readily actionable projects that new researchers and engineers can execute to immediately reduce risk. — [*Source: 80,000 Hours*](https://80000hours.org/?ref=antoinebuteau.com)

## Part 3: Effective Altruism & GiveWell

1. **On the origin of GiveWell:** GiveWell was founded out of frustration with the lack of transparent, evidence-backed information available to everyday donors trying to maximize their impact. — [*Source: GiveWell*](https://www.givewell.org/?ref=antoinebuteau.com)
2. **On quantifiable outcomes:** The impact of a charitable donation is not merely a matter of good intentions, but of measurable, cost-effective outcomes that genuinely improve lives. — [*Source: GiveWell*](https://www.givewell.org/?ref=antoinebuteau.com)
3. **On sure bets in charity:** For individual donors, it often makes the most sense to fund highly proven, heavily vetted interventions like malaria nets or antimalarial medication. — [*Source: GiveWell*](https://www.givewell.org/?ref=antoinebuteau.com)
4. **On the limitations of good intentions:** Without rigorous evaluation, even well-meaning and heavily funded philanthropic programs can fail to produce positive results. — [*Source: GiveWell*](https://www.givewell.org/?ref=antoinebuteau.com)
5. **On transparency in philanthropy:** Charities should be willing to share detailed data, admit failures, and openly discuss their weaknesses in order to improve resource allocation. — [*Source: GiveWell*](https://www.givewell.org/?ref=antoinebuteau.com)
6. **On rigorous evaluation:** Finding true cost-effectiveness requires analyzing counterfactual impact, understanding what would have happened if the organization didn't exist, rather than just overhead ratios. — [*Source: GiveWell*](https://www.givewell.org/?ref=antoinebuteau.com)
7. **On the burden of proof:** The burden of proof lies on charitable organizations to demonstrate that their interventions work, rather than on donors to prove they do not. — [*Source: GiveWell*](https://www.givewell.org/?ref=antoinebuteau.com)
8. **On comparing interventions:** It is entirely possible for one charity to be hundreds or thousands of times more effective at saving lives per dollar than an average charity. — [*Source: GiveWell*](https://www.givewell.org/?ref=antoinebuteau.com)
9. **On room for more funding:** A key metric for recommending a charity is assessing whether they can actually use additional marginal donations productively without diminishing returns. — [*Source: GiveWell*](https://www.givewell.org/?ref=antoinebuteau.com)
10. **On evidence-based giving:** Philanthropy should be treated with the same intellectual rigor and demand for evidence as scientific research or financial investment. — [*Source: GiveWell*](https://www.givewell.org/?ref=antoinebuteau.com)

## Part 4: Open Philanthropy & Hits-Based Giving

1. **On transitioning from sure bets:** While GiveWell focuses on proven interventions, the greatest potential for large-scale philanthropy often lies in funding high-risk, high-reward initiatives. — [*Source: Open Philanthropy*](https://www.openphilanthropy.org/?ref=antoinebuteau.com)
2. **On hits-based philanthropy:** Philanthropists should act more like venture capitalists, expecting many grants to fail but relying on a few massive successes to drive overall impact. — [*Source: Open Philanthropy*](https://www.openphilanthropy.org/?ref=antoinebuteau.com)
3. **On taking risky bets:** Large foundations and individuals with vast resources are uniquely positioned to take on risks that governments and traditional institutions typically avoid. — [*Source: Open Philanthropy*](https://www.openphilanthropy.org/?ref=antoinebuteau.com)
4. **On historical precedents:** Historical examples like the Green Revolution demonstrate how targeted, high-risk scientific funding can yield asymmetric benefits for humanity. — [*Source: Open Philanthropy*](https://www.openphilanthropy.org/?ref=antoinebuteau.com)
5. **On the value of failure:** A philanthropic portfolio with a zero percent failure rate is a sign of extreme risk aversion and missed opportunities for greater impact. — [*Source: Open Philanthropy*](https://www.openphilanthropy.org/?ref=antoinebuteau.com)
6. **On long-termist grantmaking:** Expanding the scope of giving to include risks facing future generations, such as pandemics and unaligned AI, is a logical extension of valuing all human lives equally. — [*Source: Open Philanthropy*](https://www.openphilanthropy.org/?ref=antoinebuteau.com)
7. **On worldview diversification:** When facing radically different moral frameworks, institutions should allocate resources across multiple worldviews rather than placing all their bets on a single ethical theory. — [*Source: Open Philanthropy*](https://www.openphilanthropy.org/?ref=antoinebuteau.com)
8. **On funding neglected causes:** The most cost-effective opportunities are often found in areas that are highly important and tractable, but currently ignored by mainstream funding sources. — [*Source: Open Philanthropy*](https://www.openphilanthropy.org/?ref=antoinebuteau.com)
9. **On expected value:** Decision-making in hits-based giving relies heavily on calculating the expected value of an intervention, multiplying the potential massive upside by its probability of success. — [*Source: Open Philanthropy*](https://www.openphilanthropy.org/?ref=antoinebuteau.com)

## Part 5: Navigating Wicked Problems

1. **On defining wicked problems:** A wicked problem is vaguely defined, lacks clear goals, and offers no stable rules or objective metrics for measuring success. — [*Source: Cold Takes*](https://www.cold-takes.com/?ref=antoinebuteau.com)
2. **On the mental health gauntlet:** Working intensely on poorly defined, high-stakes issues is exceptionally taxing and can serve as a severe mental health gauntlet. — [*Source: Cold Takes*](https://www.cold-takes.com/?ref=antoinebuteau.com)
3. **On deciding when to stop researching:** When investigating complex topics, the hardest challenge is often deciding how deep to go into any given subtopic, as there is never an objectively correct stopping point. — [*Source: Cold Takes*](https://www.cold-takes.com/?ref=antoinebuteau.com)
4. **On changing one's mind:** Frequently changing your mind about the core goal and scope of a project is not a failure, but a difficult and necessary part of the research process. — [*Source: Cold Takes*](https://www.cold-takes.com/?ref=antoinebuteau.com)
5. **On the vice of laziness:** Knowing when to stop trying to perfectly resolve a difficult question and instead shifting the focus to something more manageable is a highly useful skill. — [*Source: Cold Takes*](https://www.cold-takes.com/?ref=antoinebuteau.com)
6. **On the vice of impatience:** When lost in the weeds of research, you must regularly interrupt yourself to ask if the current line of inquiry is still relevant to the larger objective. — [*Source: Cold Takes*](https://www.cold-takes.com/?ref=antoinebuteau.com)
7. **On the vice of hubris:** You must maintain confidence in your own judgment even when navigating complex, expert-level topics where established consensus might be lacking. — [*Source: Cold Takes*](https://www.cold-takes.com/?ref=antoinebuteau.com)
8. **On self-preservation:** Prioritizing your own mental health and setting boundaries is necessary to sustain long-term engagement with difficult, open-ended problems. — [*Source: Cold Takes*](https://www.cold-takes.com/?ref=antoinebuteau.com)
9. **On learning by writing:** Writing is an effective technique for understanding wicked problems, but it must be combined with active tinkering and testing to ensure arguments map to reality. — [*Source: Cold Takes*](https://www.cold-takes.com/?ref=antoinebuteau.com)

## Part 6: Epistemology & Reasoning

1. **On minimal-trust investigations:** When exploring new fields, it is often necessary to conduct independent, minimal-trust investigations to build your own understanding rather than solely relying on credentialed experts. — [*Source: Cold Takes*](https://www.cold-takes.com/?ref=antoinebuteau.com)
2. **On expert consensus:** While expert consensus is valuable, deferring blindly to authority can be dangerous in domains where the problem is entirely unprecedented. — [*Source: Cold Takes*](https://www.cold-takes.com/?ref=antoinebuteau.com)
3. **On the value of tinkering:** Abstract reasoning should be continuously grounded by building, testing, and seeking concrete feedback in the real world. — [*Source: Cold Takes*](https://www.cold-takes.com/?ref=antoinebuteau.com)
4. **On forecasting accuracy:** Rigorous attempts to forecast the future, even if highly uncertain, are preferable to making implicit, unexamined assumptions about what will happen next. — [*Source: Open Philanthropy*](https://www.openphilanthropy.org/?ref=antoinebuteau.com)
5. **On the limits of historical analogies:** Comparing future technological transitions to past events is useful, but we must recognize when a new development fundamentally breaks historical patterns. — [*Source: Cold Takes*](https://www.cold-takes.com/?ref=antoinebuteau.com)
6. **On expected value calculations:** Expected value is a necessary tool for prioritization, but its outputs should be treated as rough guides rather than precise mathematical truths. — [*Source: Open Philanthropy*](https://www.openphilanthropy.org/?ref=antoinebuteau.com)
7. **On acknowledging uncertainty:** Intellectual honesty requires clearly communicating the large error bars and deep uncertainties surrounding long-term forecasts. — [*Source: Cold Takes*](https://www.cold-takes.com/?ref=antoinebuteau.com)
8. **On epistemic humility:** We should remain deeply suspicious of our own confidence, recognizing how often intelligent people have been spectacularly wrong about the future. — [*Source: Cold Takes*](https://www.cold-takes.com/?ref=antoinebuteau.com)
9. **On framing complex arguments:** Breaking down massive, unintuitive conclusions into step-by-step logical premises makes it easier to identify exact points of disagreement. — [*Source: Cold Takes*](https://www.cold-takes.com/?ref=antoinebuteau.com)

## Part 7: Future-Proof Ethics & Morality

1. **On common sense ethics:** Relying entirely on everyday ethical intuition is risky, as historical norms often look abhorrent in hindsight. — [*Source: Cold Takes*](https://www.cold-takes.com/?ref=antoinebuteau.com)
2. **On historical moral atrocities:** Recognizing that past societies comfortably engaged in terrible practices should make us actively search for our own current moral blind spots. — [*Source: Cold Takes*](https://www.cold-takes.com/?ref=antoinebuteau.com)
3. **On expanding the moral circle:** Moral progress involves systematically expanding our circle of concern beyond our immediate tribe to include all humans, animals, and future generations. — [*Source: EA Forum*](https://forum.effectivealtruism.org/?ref=antoinebuteau.com)
4. **On non-human animal welfare:** Factory farming represents a massive, ongoing moral catastrophe that requires urgent philanthropic and societal attention. — [*Source: Open Philanthropy*](https://www.openphilanthropy.org/?ref=antoinebuteau.com)
5. **On the moral weight of future generations:** Because future generations cannot advocate for themselves or participate in current markets, they are systematically disenfranchised in today's decision-making. — [*Source: 80,000 Hours*](https://80000hours.org/?ref=antoinebuteau.com)
6. **On rigorous ethical frameworks:** We should seek out systematic frameworks for ethics to help us avoid repeating the tragic moral errors of the past. — [*Source: Cold Takes*](https://www.cold-takes.com/?ref=antoinebuteau.com)
7. **On avoiding the mistakes of the past:** Future-proofing our ethics requires actively challenging our strongest intuitions and remaining open to radically different moral conclusions. — [*Source: Cold Takes*](https://www.cold-takes.com/?ref=antoinebuteau.com)
8. **On extreme poverty vs long-term risks:** Balancing the immediate, visceral suffering of extreme poverty against the abstract, probabilistic risks to the distant future is one of the hardest challenges in cause prioritization. — [*Source: Open Philanthropy*](https://www.openphilanthropy.org/?ref=antoinebuteau.com)
9. **On the strange implications of morality:** Taking morality seriously often forces us into conclusions that seem highly counterintuitive or socially awkward by contemporary standards. — [*Source: Cold Takes*](https://www.cold-takes.com/?ref=antoinebuteau.com)

## Part 8: Career Capital & Building Aptitudes

1. **On transferable skills:** In an uncertain world, building versatile, highly transferable skills is often a safer and more impactful strategy than prematurely narrowing your focus. — [*Source: 80,000 Hours*](https://80000hours.org/?ref=antoinebuteau.com)
2. **On kicking ass in early career:** Early in your career, focus intensely on roles where you can demonstrate high competence, deliver results, and learn how to function at a professional standard. — [*Source: 80,000 Hours*](https://80000hours.org/?ref=antoinebuteau.com)
3. **On generalist vs specialist paths:** While specialized knowledge is vital, highly capable generalists who can synthesize information and navigate wicked problems are incredibly rare and valuable. — [*Source: 80,000 Hours*](https://80000hours.org/?ref=antoinebuteau.com)
4. **On the value of hard work:** Producing a high volume of quality output is one of the most reliable ways to build career capital and discover your true comparative advantages. — [*Source: 80,000 Hours*](https://80000hours.org/?ref=antoinebuteau.com)
5. **On finding comparative advantage:** Impact is maximized by finding the intersection of an important problem and your unique personal strengths. — [*Source: 80,000 Hours*](https://80000hours.org/?ref=antoinebuteau.com)
6. **On shifting focus when necessary:** You should be willing to pivot your entire career trajectory if new evidence suggests that a different cause area has become significantly more urgent. — [*Source: 80,000 Hours*](https://80000hours.org/?ref=antoinebuteau.com)
7. **On building a track record:** Earning the trust to take on independent, high-impact work requires first building a rock-solid track record of reliability in structured environments. — [*Source: 80,000 Hours*](https://80000hours.org/?ref=antoinebuteau.com)
8. **On the importance of operational roles:** Excellent operations, management, and administrative skills are massive bottlenecks for impactful organizations, and are just as necessary as research talent. — [*Source: 80,000 Hours*](https://80000hours.org/?ref=antoinebuteau.com)
9. **On career flexibility:** Given the potential for rapid technological transformation, maintaining the flexibility to adapt to radically altered professional landscapes is essential. — [*Source: 80,000 Hours*](https://80000hours.org/?ref=antoinebuteau.com)