Elad Gil: How to Spot a Billion-Dollar Startup Before the Rest of the World
Elad Gil, an investor behind Airbnb, Stripe, and Coinbase, shares his mental playbook on startups, talent, decision-making, and AI. He discusses the importance of geographic clusters for innovation, common self-inflicted wounds that kill companies, and how to scale a CEO.
Deep Dive Analysis
23 Topic Outline
Investing Philosophy: Market Opportunity vs. Team
Identifying Product Market Fit and Outlier Teams
The Phenomenon of Geographic Innovation Clusters
Remote Work's Impact on Innovation and Collaboration
Y Combinator's Role in Founder Development
Evolution of AI Waves: Researchers to Applications
AI as Units of Cognition and Future Capabilities
Short-Term and Long-Term Fears Regarding AI
AI's Current and Future Problem-Solving Beyond Human Capacity
Challenges of Data Quality in Scientific Research
Prompt Engineering and Interacting with AI Models
Bottlenecks and Scaling in AI Development
Incumbents vs. Startups in the AI Revolution
The Strategic Importance of Open Source AI
Regulation and Safety Concerns in AI Development
Societal Risk Aversion and its Negative Implications
Common Self-Inflicted Wounds That Kill Companies
Scaling the CEO and Avoiding Conventional Wisdom
Workplace Culture: Team vs. Family Mindset
Patterns Among Outlier CEOs and Drivers of Longevity
The Power of Compounding in Careers and Information Flow
Investing in Defense Technology: The Anduril Case Study
Future of Warfare: Drones, Cyber, and Autonomous Systems
8 Key Concepts
Product Market Fit
This refers to whether there is a strong demand for what a company is building. For consumer businesses, it's indicated by organic growth rate and retention; for B2B, by growth, adoption, and metrics like net dollar retention. For pre-product, it relies on belief in customer need.
Innovation Clusters
These are geographic areas where groups of young people aggregate, find each other, and collaborate towards common goals, leading to major movements in various fields like art, literature, or technology. Silicon Valley is a prime example for tech, where talent self-aggregates and supports each other.
Transformer Architecture
A specific paper published by Google in 2017 that describes a neural network architecture. This architecture is credited with spawning the current wave of AI advancements, enabling new capabilities not possible with older deep learning technologies.
Units of Cognition
A future way of thinking about AI, where instead of selling tools, AI will be perceived as selling 'bits of person time' or 'person equivalent' to perform tasks. This implies hiring 'bot programmers' or 'AI accountants' by renting time from these cognitive units.
Self-Play (AI)
A method of training AI models where the model learns by playing against itself, rather than being shown pre-existing human data. This approach requires clear rules and a utility function (like winning a game) to provide rapid feedback for training and iteration, leading to creative and superhuman performance.
Prompt Engineering
The practice of crafting specific and detailed instructions for AI models to achieve desired outcomes. It often involves breaking down complex tasks into steps, adding specificity, and incorporating safeguards or 'test cases' to ensure accuracy and prevent errors, similar to guiding a smart intern.
Cost-Plus Business Model
A business model prevalent in the defense industry where contractors are paid a percentage (e.g., 5-12%) on top of their total costs, including labor and components. This model creates incentives for cost overruns and delays, as it directly increases the contractor's profit, leading to inefficiencies and inflated prices.
Missionary vs. Mercenary
A framework for understanding what drives individuals, particularly founders. Mercenaries are primarily driven by money and survival, especially when young. Missionaries are driven by a broader purpose and impact, often emerging in the middle phase of a career. Later in life, some become 'artists,' doing things for the love of the craft.
12 Questions Answered
Elad Gil prioritizes market opportunities and product market fit first, asking if there's a real need and differentiation for what's being built. Only then does he evaluate the team's exceptional qualities and potential for growth, a reversal of how many early-stage investors operate.
An outlier tech team often embodies the 'Apple framework' of Jobs, Wozniak, and Cook: at least one person skilled at vision-setting, sales, and negotiation; someone uniquely capable of building and hacking things together; and eventually, someone who can scale the company.
Innovation clusters form because young, talented people aggregate in specific cities, find each other, and collaborate, sharing ideas, learning new methods, and benefiting from concentrated funding and experienced mentors. This pattern has been observed across various historical movements, not just tech.
Remote work is generally not conducive to innovation, as it hinders the informal idea sharing, collaboration, and self-aggregation of talent that occurs in physical clusters. While tech is often thought to be suited for remote work, other industries like film or finance also cluster despite remote possibilities.
The Transformer Architecture, a paper from Google in 2017, is the foundational technology that spawned the current wave of AI, including large language models like ChatGPT. It enabled a step-function increase in capabilities compared to prior deep learning technologies.
In the short term, Elad Gil fears that over-regulation and constraint could 'strangle the golden goose' of AI, preventing its potential for global advancements in health and education. In the long term, concerns revolve around questions of AI sentience, new life forms, and potential species competition, especially if combined with robotics.
AI has already surpassed human capabilities in many areas, such as gaming (Go, chess, poker, diplomacy). In scientific fields like biology, AI is making advancements (e.g., protein folding), but its ability to solve complex mathematical or biological problems where data is messy or incomplete is still developing, requiring clean datasets and appropriate model architectures.
For early-stage companies, common self-inflicted wounds include founders fighting and the team blowing up, or running out of money before achieving product market fit. Later, companies can fail by becoming too competitor-centric instead of customer-centric, losing differentiation, or engaging in actions that hurt themselves more than competitors.
Successful CEOs scale by recognizing what they don't need to reinvent (like sales processes) and hiring effective people to fill those roles. They also build an organizational structure that complements their unique quirks and obsessions, rather than adhering strictly to conventional wisdom or generic team structures.
A healthy workplace culture prioritizes the company's mission and objectives, focusing on performance and effective collaboration. It views the organization as a 'team' rather than a 'family,' meaning individuals are expected to bring their professional selves to work and contribute towards shared goals, rather than tolerating unproductive behaviors.
Outlier CEOs are typically incredibly smart and insightful. They often fall into archetypes: hyper-focused individuals dedicated solely to their company, or 'polymathic' individuals with broad interests who dive deep into various fields while also running their companies. Many are driven by a blend of impact, interestingness, and purpose beyond just money.
To improve information flow and trajectory, individuals should consider moving to geographic clusters relevant to their desired field to be immersed in the flow of ideas and talent. The internet also helps by creating online 'gold medalist communities' where like-minded people can connect and share information, fostering a different environment for finding connections.
45 Actionable Insights
1. Define Your Core Utility Function
Clearly define your personal ‘utility function’ – what truly drives and fulfills you beyond just money – to ensure long-term happiness and purpose, as solely financial motivation often leads to misery after achieving wealth.
2. Evolve from Mercenary to Artist
Embrace a career progression that starts with mercenary drive (for survival and success), transitions to a missionary focus (broader purpose), and culminates in an ‘artist’ mindset (doing it for the love of the craft), as this path leads to long-term happiness and significant impact.
3. Cultivate Polymathic Interests and Early Success
To achieve long-term relevance and impact, cultivate a polymathic approach with diverse interests, be driven by a blend of motivations beyond just money (e.g., impact, interestingness), and strive for early success to build a platform for compounding career growth and access to valuable information flow.
4. Prioritize Product-Market Fit in Investing
When evaluating a startup or building one, first assess if there’s a strong demand and differentiation for the product (product-market fit) before evaluating the team, as even amazing teams can fail in bad markets, while mediocre teams can succeed in good ones.
5. Relocate to Industry Clusters for Success
To maximize your chances of being at the top of your game in a specific industry, relocate to its primary geographic cluster (e.g., Hollywood for movies, Silicon Valley for tech) to leverage the concentrated talent and information flow, rather than making it harder by working remotely from anywhere.
6. Gain Early Access to Emerging Tech
To identify future opportunities and innovations, position yourself to gain early, firsthand access to emerging technologies and their internal usage, as those closest to the tech often see the future before it becomes widely distributed.
7. Develop Superior Information Interpretation
To gain an information advantage, focus not just on accessing unique data, but critically on developing superior methods to interpret publicly available information, including having the right tools, filters, and the ability to make intuitive leaps that others miss.
8. Stay Customer-Centric, Not Competitor-Centric
To avoid self-inflicted wounds, remain customer-centric rather than becoming overly focused on competitors, as excessive competitor focus can lead to losing differentiation, neglecting customer needs, and undertaking actions that don’t benefit your company.
9. Treat Your Company as Performance Team
View your company as a performance-oriented team with a specific mission, rather than a family, to optimize for collective performance and ensure all members are aligned and contributing effectively towards shared goals.
10. Foster Independence and Risk-Taking
Resist societal trends that reduce independence, agency, and risk-taking, especially in children, as an overemphasis on safety and fragility can have negative long-term implications for mental health and the ability to function independently.
11. Embrace Peer Pressure for Rapid Shipping
To accelerate product development and customer acquisition, immerse yourself in environments (like YC batches) that create peer pressure and set strict timelines, fostering a mindset of shipping fast and constantly seeking customer feedback.
12. Build a Balanced Startup Founding Team
For an early tech team, ensure you have at least one person skilled in vision-setting and sales (like Jobs), one person exceptionally good at building (like Wozniak), and plan for someone to scale the company as it grows (like Cook) to cover essential roles.
13. Choose Successors with Founder Mindset
When planning CEO succession, prioritize finding a leader who possesses a product-founder mentality, capable of innovation and driving new changes, rather than simply promoting a complementary operational lieutenant, to prevent company decay after a founder’s golden age.
14. Design Org Structure to CEO’s Needs
Tailor the organizational structure of a large company to specifically complement the CEO’s unique needs, quirks, and obsessions, rather than adopting a generic structure, to optimize for their effectiveness and drive.
15. Avoid Unnecessary Reinvention as CEO
As an innovative CEO, recognize areas that are well-established (like sales processes) and avoid the urge to reinvent them; instead, hire experienced professionals to manage these functions, allowing you to focus innovation where it’s truly needed.
16. Hire Complementary, Effective Team Members
When scaling yourself as a leader, focus on hiring individuals who are highly effective in their roles and possess skills that complement your own, rather than trying to do everything yourself.
17. Bring Your Professional Self to Work
Focus on bringing your professional self to work, emphasizing effectiveness and collaborative behavior, rather than your ‘whole self’ with all personal morals and values that may not be appropriate for a diverse work environment.
18. Prioritize In-Person for Innovation
For companies focused on innovation, especially young ones, prioritize in-person work environments over remote work, as remote setups are generally less conducive to spontaneous collaboration and idea sharing unless specifically designed for online collaboration.
19. Implement Robust Processes for Remote Work
If committing to a remote-first company model, implement comprehensive, well-documented processes for every aspect, from salary bands to communication protocols, to ensure efficiency and effective collaboration, as exemplified by GitLab.
20. Evaluate Remote Policy’s Impact on Commitment
CEOs should critically evaluate their remote work policies, recognizing that enforcing in-person attendance can implicitly prioritize employees who are more committed to the company’s mission over other personal priorities, and make a judgment call on its importance.
21. Re-evaluate Big Tech Staffing Levels
For big tech companies, consider a drastic re-evaluation of staffing levels and productivity, as many may be overstaffed with employees not working very hard, suggesting potential for significant cuts without compromising output.
22. Reframe AI as Units of Cognition
Shift your mindset to view AI as ‘units of cognition’ or ‘person-equivalent time’ that can be rented or hired to perform white-collar work, enabling you to effectively scale capabilities by leveraging AI for tasks like programming or accounting.
23. Advocate for Unconstrained AI Development
Support and advocate for the unconstrained development of AI, as its potential for global advancements in health, education, and other fundamental areas is immense and should not be stifled by short-sighted regulation.
24. Use Specificity and Checks in AI Prompts
When prompting AI, be highly specific in your requests and incorporate explicit checks or ’test cases’ within the prompt (e.g., ‘double check that these two things are true’) to reduce errors and improve the reliability of the output.
25. Leverage AI for Information Synthesis
Use AI for practical tasks like synthesizing information (e.g., identifying key people at a conference and pulling their backgrounds), coding, and generating regular expressions to extract specific data, enhancing productivity in various domains.
26. Break Down Complex AI Tasks
When prompting AI for complex tasks, break the task down into a series of smaller, sequential steps and chain the model by feeding the output of each step as input for the next, similar to how an agent would process information.
27. Focus AI on Rapid Feedback Domains
Prioritize applying AI to domains that offer clear, testable outputs and rapid feedback loops, such as coding, as this allows for quick iteration and continuous training of the system.
28. Utilize Self-Play for AI Training
When developing AI, especially in domains with clear rules and a defined utility function (like winning a game), implement self-play training where the AI generates its own data and receives rapid feedback for continuous improvement.
29. Prioritize Clean Data for AI in Biology
When applying AI to complex fields like biology, focus on creating high-fidelity, clean datasets for training models, as this foundational work is crucial for enabling AI to make accurate breakthroughs and clean up existing data.
30. Develop AI Fraud/Plagiarism Detectors
Consider developing AI-powered fraud or plagiarism detectors for both liberal arts and sciences, as there’s a significant opportunity to uncover widespread issues and improve academic and scientific integrity.
31. Support Open Source AI for Customization
Advocate for and utilize open-source AI models because they level the playing field, make technology globally accessible, and allow users to customize or retrain models to reflect specific cultural norms or values, avoiding imposed biases.
32. Define ‘Safety’ Clearly in Discussions
When discussing ‘safety’ (e.g., AI safety, societal safety), clearly define what specific type of safety is being addressed (digital, physical, existential) and articulate why it constitutes a real concern, to avoid conflation and unproductive arguments.
33. Prevent Founder Conflict and Achieve Product-Market Fit
For early-stage companies, actively manage founder relationships to prevent internal conflict and ensure rapid progress towards achieving product-market fit, as these are the primary causes of early company failure.
34. Measure Product-Market Fit with Metrics
For consumer businesses, track organic growth rate and retention to see if people are using the product frequently; for B2B products, monitor growth rate, adoption, and metrics like Net Dollar Retention (NDR) to gauge demand.
35. Join Programs to Access Networks
If you are a talented individual outside of core industry networks, seek out programs or accelerators (like Y Combinator) that can plug you into established networks to help you succeed.
36. Seek Lineage from Top Labs/Groups
To maximize your potential for breakthroughs and success in any field, strive to join or establish a lineage with the handful of top labs, groups, or institutions that consistently drive most of the progress and produce successful individuals.
37. Reconsider Equal Co-Founder Myth
Challenge the conventional wisdom that you must have an equal co-founder; instead, consider solo founding or an unequal founder structure, as many of the biggest startup successes have originated from these models.
38. Contextualize Negative Founder References
When conducting reference checks on founders, contextualize negative feedback; unless it points to ethical issues, it might be neutral, as product-market fit can trump individual founder traits, and performance can drastically improve when an individual is fully responsible for their own venture.
39. Identify Opportunities in Neglected Sectors
Look for startup opportunities in sectors that large, incumbent companies are neglecting or deeming unpopular (e.g., defense when Google exited Maven), as this creates a significant opening for new ventures.
40. Identify ‘Why Now’ Tech Shifts
When entering an established industry, identify a ‘why now’ technological shift that incumbents cannot easily adopt or tack on, creating a unique competitive advantage for a new company.
41. Build Broad Product Portfolio as Prime
To succeed in industries dominated by large players (like defense), build a sufficiently broad product portfolio from day one to position your company to become a ‘prime’ contractor or major player.
42. Secure Sufficient Funding for Long Timelines
For ventures with inherently long development or sales cycles (e.g., defense contracts), ensure you raise enough capital to sustain operations and endure extended timelines required to achieve major milestones.
43. Innovate Business Models for Efficiency
Challenge and innovate traditional, inefficient business models (like cost-plus in defense) by proposing alternatives that align incentives for both the provider and the customer, leading to better value and efficiency for all parties.
44. Embrace Drone-Based Defense Systems
For military and defense strategists, recognize and embrace the inevitable shift towards highly distributed, drone-based systems, as they offer advantages in speed, cost, and operational capabilities beyond human limitations.
45. Maintain Human Oversight in Lethal AI
When developing or deploying AI in defense, ensure that a human operator always maintains control over decisions involving lethal force, rather than ceding full autonomy to the system, to prevent unintended harm and maintain ethical boundaries.
8 Key Quotes
AI is the only market where the more I learn, the less I know.
Elad Gil
AI is dramatically under-hyped because most enterprises have not done anything in it. And that's where all the money is, all the changes, all the impact is, all the jobs, everything.
Elad Gil
The people that I know who have been very successful or driven solely by money end up miserable. Because they have money, and then what?
Elad Gil
Remote work is generally not great for innovation unless you're truly in an online collaborative environment.
Elad Gil
I think in a couple of years we'll start thinking about it as we're selling units of cognition.
Elad Gil
The future is here is just not equally distributed.
Elad Gil
You actually should bring your professional self to work. You should bring the person who's going to be effective in a work environment and can work with all sorts of diverse people and be effective and doesn't bring all their morals and values and everything else in the workplace that don't have a place in the workplace.
Elad Gil
The thing I continue to underappreciate is the power of compounding. And you see that in investing and financial markets, but you also see that in people's careers and impact.
Elad Gil