Inside Early Facebook with Andrew Bosworth: Culture, Ads, and AR Glasses

Key Concepts:

• Chaotic but Fun Startup Culture: Early Facebook was characterized by chaos, naivete, and fun, fostering camaraderie and optimism.
• Minus One Journey: The process of rediscovering and refocusing oneself, often involving reinventing one's role within a company.
• Human Intuition vs. Machine Learning: The balance between human expertise and automated systems, highlighting the value of human intuition in certain contexts.
• Culture of Ownership: A work environment where individuals take personal responsibility and care deeply about the success of the company.
• Year of Greatness: The idea that certain periods offer disproportionate opportunities for change and impact.
• Software as the Limiter in Robotics: The shift in robotics where software development, particularly AI and machine learning, is the primary constraint.
• Skeptic in Charge: The strategy of placing someone with a critical perspective in a leadership role to challenge existing assumptions and drive innovation.
• Cold Start Algorithm: A method of gathering information and identifying key problems by talking to people and asking questions.
• Importance of Inventory (Ads): The concept that having a large volume of ads, even if initially cheap, leads to increased competition and higher prices.
• Embracing Different Working Styles: Recognizing and accommodating diverse work preferences and capabilities within a team.
• Subjective Taste: The recognition that some individuals possess better judgment and discernment in design, marketing, and other areas.
• Connecting Technology and National Defense: The importance of collaboration between the technology sector and the military to maintain competitiveness and security.

1. Early Facebook Culture and Joining the Company:

• Andrew Bosworth (Boz) joined Facebook on day two as user 1,678.
• He was a teaching fellow for Mark Zuckerberg's AI class at Harvard.
• Dana Hayes recruited him from Microsoft, initially dismissed as unbelievable.
• Early Facebook culture was described as "chaotic," "super naive," and "fun."
• The story of "Richie plugging the internet into itself" illustrates the chaotic environment. A misconfigured server and firewall caused a network-wide ping flood due to an Ethernet cable plugged into itself.
• Startups should be fun, with a sense of camaraderie and optimism.
• Early Facebook felt like "riding lightning," with a belief in the company's potential.

2. Bosworth's Minus One Journey and Newsfeed:

• Bosworth constantly reinvented himself within Facebook, likening his career to the "Karate Kid" analogy of "painting fences and waxing cars."
• He worked on Newsfeed, early integrity efforts, and anti-spam measures.
• Newsfeed aimed to aggregate changes from profiles into one place, like a personal newspaper.
• He manually created ranking curves for event stories and photo albums, using mathematical functions to match the curves.
• Mark Zuckerberg would provide feedback, such as reducing the number of basketball stories.
• Bosworth woke up every four hours for two years to monitor and combat scrapers and spammers, manually intervening in the database.
• He later realized the importance of working with and leading teams, not just being a solo technologist.

3. Human Intuition vs. Machine Learning in Newsfeed:

• Bosworth initially hand-tuned weights for Newsfeed's story composition.
• A machine learning team spent 18 months trying to automate this process but couldn't surpass Bosworth's hand-tuned metrics.
• This highlighted the power of human intuition and the limitations of machine learning at the time.
• Google Search only fully switched to machine learning a few years ago, previously using human-tuned algorithms to refine machine learning outputs.

4. Culture of Ownership and Competitiveness:

• Early Facebook had a culture of total ownership, where individuals cared deeply and took personal responsibility.
• No one had to tell employees to care more; they were intrinsically motivated.
• Bosworth felt personally at war with spammers and was determined to win.
• Facebook's culture is a microcosm of Mark Zuckerberg's DNA, emphasizing competitiveness and a "failure is not an option" mentality.
• Zuckerberg leads by example, never asking anyone to work harder than he is willing to.
• Even high-level executives like Ruchi are willing to perform basic tasks, demonstrating that no job is too big or small.
• Reed Hastings' story about the Netflix CEO washing mugs illustrates a similar principle.

5. Current Role as CTO and the "Year of Greatness":

• Bosworth feels fortunate to be in his current role, especially during a time of significant technological change.
• He believes that certain periods offer disproportionate opportunities for impact.
• The current era with AI and hardware innovation is similar to the early days of the internet.
• He wrote an internal memo called "Year of Greatness," emphasizing the importance of working hard during this unique time.
• Every year of one's career is not equivalent; some years offer more potential for compounding returns.
• Meta has the best AI wearable in the market and needs to double down to maintain its lead.

6. Robotics and the Shift to Software as the Limiter:

• Hardware is no longer the primary constraint in robotics; software is.
• Even older robots could perform complex tasks when teleoperated, indicating the hardware's capability.
• Modern electromechanical actuators have improved hardware, but software limitations remain.
• Locomotion through reinforcement learning has made progress, but robots still struggle with interesting and useful tasks.
• The focus is now on grippers and hands, where machine learning is crucial.
• Conventional roboticists relied on heuristics, which were insufficient.
• Robotics was late to adopt modern machine learning, only recently embracing large language models and world models.
• The belief that software will arrive is driving current hardware development.

7. Transition to Mobile Ads and the "Cold Start Algorithm":

• Bosworth was initially a product infrastructure guy, not an ads guy.
• Mark Zuckerberg often puts skeptics in charge of things.
• Bosworth used a "cold start algorithm" to learn about the ads business by talking to people and asking questions.
• He conducted 25-minute meetings, dedicating 20 minutes to listening and 5 minutes to asking for further contacts.
• Knowing nothing was an asset, allowing him to approach the problem with fresh eyes.
• Facebook made most of its money on the website at the time, despite having the most popular mobile app.
• The problem was primarily a market problem: needing more ads.
• Bosworth convinced Zuckerberg to increase the volume of ads, even if initially cheap, to drive competition and higher prices.
• He focused on strategic and business problems, rather than just technical ones.
• Self-imposed limits based on identity can hinder progress.

8. Glasses and Organizational/Cultural Change:

• Bosworth was initially skeptical about AR/VR glasses.
• Zuckerberg used a "Jedi mind trick" to get Bosworth on board, asking him what would need to be true for it to be a good idea.
• This led Bosworth to develop a roadmap and strategy, becoming enthusiastic about the project.
• Facebook's previous attempt at building a phone felt different from the glasses project.
• Building hardware and working with an extensive partner ecosystem required significant organizational and cultural change.
• Zuckerberg felt "boxed in" by mobile platforms and wanted to invent the next hardware platform.
• Facebook is not a homogeneous entity; different teams operate with different styles and cadences.
• Chris Cox runs family of apps, and Javier Olivan runs the operational side.
• Bosworth's group operates more independently, with longer cycle times and more trust from Zuckerberg.
• Mark Zuckerberg is skilled at understanding people's strengths and putting them in positions to maximize those strengths.

9. Building for the Long Term and Subjective Taste:

• Bosworth's teams build just as fast as others, but their deliverables have longer timelines.
• Assessing progress in the middle of a technological cycle is a crucial skill.
• The market cannot immediately validate their features, so they must proxy the market.
• User research, prototyping, and internal communities are used to gather feedback.
• Taste exists, and people with a history of success in bringing new technologies to market are valuable.
• The loss rate is not significantly different from the web, but they have activated their ability to assess the potential of ideas.
• Startups should paint a future where underlying core bets pay off.
• It is important to assess whether someone has good subjective taste in design, marketing, and code quality.
• The hardware industry already had systems for validating industrial design and comfort.
• Getting software people to think outside the mobile phone frame and embrace long-term planning has been more challenging.

10. Lessons from the Army Reserve:

• There is a universality to leadership principles; the army is just managing people.
• It is a trap to think of oneself as being inside a system; one must be willing to break the system.
• The personnel in the army are brilliant and capable but lack the necessary tools.
• The technology sector and national defense sector should be better connected.
• The enmity felt in Silicon Valley towards the military-industrial complex was misplaced and harmful.
• Companies should be open to pursuing contracts with the US government.

11. Recommendations:

• Read "Insanely Great" by Stephen Levy.
• Read about the computing revolution of the 1960s and 70s.
• Listen to the Lockheed Martin podcast and Acquire.

12. Conclusion:

The conversation provides valuable insights into the culture of early Facebook, the importance of adaptability and continuous learning, the balance between human intuition and machine learning, and the need for collaboration between the technology sector and national defense. Bosworth's experiences highlight the significance of embracing chaos, fostering a culture of ownership, and being willing to challenge existing assumptions to drive innovation. The discussion also emphasizes the importance of subjective taste and the need to connect with the past to shape the future of computing.