Alex Komoroske: Systems Thinking, Builders vs Gardeners, and Working In Large Organizations

Alex Komoroske discusses the transformative potential of large language models (LLMs) in AI, likening them to "magical duct tape" that can significantly enhance human creativity and collaboration rather than acting as independent agents. He emphasizes the importance of systems thinking and iterative approaches in navigating complex environments, advocating for incremental, adaptive strategies over rigid, grand plans. Komoroske also highlights the need for a new security model to safely integrate untrusted third-party code, envisioning a future where software is more fluid, personal, and empowering. His insights draw from his extensive experience at Google and his forward-looking perspective on AI's role in society.

• LLMs (Large Language Models) are compared to trained circus bears in a kitchen.
• The metaphor highlights the miraculous yet unpredictable nature of LLMs.
• LLMs lack agency and are better viewed as tools rather than autonomous agents.

"I kind of joke that LLMs which are magical and amazing are kind of like having a trained circus bear in your kitchen that makes you porridge."

• Highlights the miraculous nature of LLMs despite their unpredictability.

"It's a miracle that it works at all but like also it's a wild animal that is in your kitchen and could like ransack your kitchen or kill you."

• Emphasizes the potential risks and unpredictability associated with LLMs.

• Systems thinking involves understanding and influencing systems rather than controlling them.
• The gardening metaphor illustrates nurturing and influencing systems without direct control.
• The gardener mindset contrasts with the builder mindset, focusing on indirect influence and patience.

"One of the core things that is a commonality across a lot of the work I've done... it comes down to systems thinking, gardening systems, swarms."

• Introduces the concept of systems thinking and its relevance across various domains.

"To really wrestle with systems you have to let go and just dance with the system."

• Emphasizes the need to adapt and interact with systems rather than trying to control them.

"The gardener is just sitting there, you know, dilly-dallying and looking at stuff at his feet and looking at picking up little pebbles."

• Illustrates the patience and observational approach of the gardener mindset.

• The parable contrasts the approaches of a builder and a gardener in creating large, beautiful outcomes.
• The builder focuses on immediate, tangible actions, while the gardener nurtures potential growth over time.
• The gardener's approach is less about direct control and more about fostering conditions for growth.

"The Builder gets immediately to work and says, okay, here's some rocks I can pick up and here's the heaviest one I can carry."

• Describes the builder's immediate, action-oriented approach.

"The gardener is just sitting there, you know, dilly-dallying and looking at stuff at his feet and looking at picking up little pebbles."

• Highlights the gardener's patient and observational approach.

"The gardener understands that other things can be alive and that some of those pebbles are actually acorns."

• Explains the gardener's recognition of potential growth in seemingly insignificant elements.

• Two types of real magic in leadership: Soromon (builder) and Radagast (gardener).
• Soromon archetype involves heroic, powerful leadership with a reality distortion field.
• Radagast archetype involves nurturing and developing the greatness in others, often working behind the scenes.

"The Soromon is the typical Steve Jobs is a canonical Soromon... heroic, powerful, insightful startup leader."

• Describes the Soromon archetype and its characteristics.

"The Radagast magic works for a totally different world... loving everything and everyone around you and seeing seeds of greatness everywhere you look."

• Describes the Radagast archetype and its nurturing approach.

"This magic fundamentally, to wield it you have to let go of the idea of getting credit for what you do."

• Highlights the selfless nature of the Radagast archetype.

• Radagast leaders are often less visible and work behind the scenes to create significant impact.
• Examples include Steve Wozniak, Stuart Brand, Kevin Kelly, and Dee Hock.
• These leaders inspire and catalyze larger outcomes without seeking personal credit.

"Steve Wozniak is a great example, Stuart Brand, Kevin Kelly... these kinds of people who inspire and catalyze something much larger than themselves."

• Lists examples of Radagast leaders and their impact.

"Dee Hock was one of the people who I didn't know about before I worked at a payments company... 100% chaotic organizations like the whole his whole vibe."

• Describes Dee Hock's influence and approach as a Radagast leader.

• Understanding the complexity of organizations can lead to significant influence.
• Leaders who embrace complexity can create compounding positive outcomes.
• Effective leaders often have one foot inside and one foot outside the system to see and influence it effectively.

"People who understand the actual complexity of the organization that they're in tend to become Hermits."

• Describes the tendency of those who understand complexity to withdraw and observe.

"The people who are best positioned to have the largest leverage have one foot inside the system and one foot outside the system."

• Emphasizes the importance of being both inside and outside the system to effectively influence it.

• Mentorship is crucial for developing systems thinking and understanding complex environments.
• Mentoring provides concrete challenges and helps distill intuitive knowledge into words.
• Product management often involves systems thinking, especially in platform roles with multiple stakeholders.

"Mentoring is one of those things where it adds value directly... it has a huge impact on someone's life and career."

• Highlights the value of mentorship in developing systems thinking.

"Product management is actually that you're in much less control of your product and its usage than you actually think you are."

• Describes the illusion of control in product management and the importance of systems thinking.

"I was the PM for the open web platform for a number of times for a number of years... you've got multiple browser vendors all of them are don't really like each other that much."

• Illustrates the complexity and need for systems thinking in platform product management.

• Systems thinking involves understanding the interconnectedness and feedback loops within a system.
• Feedback loops can be long, and cause and effect can be difficult to discern, making it hard to gain confidence in systems thinking.
• Constraints within a system must be acknowledged and worked with, much like physical laws such as gravity.

"The feedback loops are so long in some cases and the cause and effect are such a black box that it's really even if you are good or are becoming good, it's difficult to gain confidence in that."

• Systems thinking is about making the right thing happen in a large organization with emerging effects that no one individual fully understands.

"You have to work with these people like this is a constraint like this is like saying well if gravity didn't exist then this would work."

• Systems thinking involves planting a diverse set of small, viable ideas (acorns) and investing energy in those that grow.
• It's important not to hold tightly to any particular plans but to respond to what works.

"You plant a whole bunch and then respond to the ones that are going in the direction you want and don't hold on too tightly to any particular plans."

• Supporting individuals with unique insights can lead to significant improvements and innovations.

"The world would be a better place if I help this person become an even more effective version of themselves."

• Constraints can be visible (normal constraints), hidden (dinosaur bones), or subtle and easily forgotten (laser beams).
• Understanding what people actually think, not just the official stance, can uncover viable solutions.

"Constraints can be normal constraints everybody can see; these are easier. They can also be like dinosaur bones; they're concrete but they're buried, and you got to unearth them."

• Engaging with people authentically can reveal hidden constraints and foster real connections.

"One I would curse, drop the f-bomb within the first 30 seconds, and two I would observe something embarrassing about the leadership team."

• Cold reading statements can facilitate interesting conversations and uncover hidden truths within an organization.
• Using these tactics requires a moral compass to ensure they are not used harmfully.

"I believe in the official strategy, and I think it's going to work, but man, are there a lot of challenges."

• Continuously asking if one's actions would be embarrassing if shown to a respected audience helps maintain ethical behavior.

"If someone were to show me a video of this in 10 years in front of a hundred people whose opinion I deeply care about, would I be embarrassed?"

• Sometimes, immediate action is necessary to maintain momentum, even if it's not the perfect action.
• Organizations need a balance of optimism (kayfabe) and realistic assessment to function effectively.

"If the organization is going to die if you don't take some action in the next step, like, take some action, man."

• Kayfabe, a term from professional wrestling, describes the necessary facade that everyone knows is fake but acts as if it's true to maintain organizational function.

"Kayfabe is one of the things that makes organizations work if every time that someone proposed a new project, a leader said we're going to do so and so someone raised their hand in front of everyone goes this will definitely not work."

• Planting seeds involves introducing ideas that can grow into significant changes when the organization is ready.
• Safe subversion means hinting at ideas without fully connecting the dots, allowing others to draw their own conclusions and feel a part of the solution.

"You connect nine of the ten dots and leave the last dot unconnected, and that way, if you connect all the ten dots, you see something subversive."

• This approach can lead to significant organizational changes without direct confrontation.

"If you get the right belly laugh at the exact right moment, you might get the organization to go on the right path."

• Ensuring that one's role or team is perceived as valuable and competent is crucial for survival in an organization.
• Allocating a portion of time to experimental and innovative projects can lead to significant breakthroughs.

"If anybody ever asks somebody else about you or your team and they go, what does so and so do again? That means you're on the cusp of death."

• A balance of routine work and deliberate experimentation can yield valuable results.

"70% of my time should be doing like normal, like just do the thing everybody thinks that we should be doing, and then 30% of the time would have been slop and random."

• A real example involves mentoring a product manager to develop a new strategy that was later adopted after a reorganization.
• The strategy was introduced subtly and gained traction when the organization was ready for change.

"Imagine an entire product area of a large organization that is working on a thing that if you ask every individual person behind the scenes and say, does this strategy work? Everyone goes, yeah, definitely not."

• This approach shows the importance of timing and subtlety in implementing significant changes.

"So this person shared this document around; it was read by a number of people, and they're like, oh yeah, that's a good idea. And nothing really happened. And then the reorg happens, and a new VP comes in."

• Writing and sharing ideas can be a way to think through concepts and gain feedback.
• The risk of sharing ideas is often lower than perceived, as unnoticed ideas have minimal impact.

"I think by talking and like writing to me as a way of thinking through ideas, and a bunch of people will talk to me in my office hours and say, you know, I aspire to blog more often."

• Focusing on the potential impact rather than the fear of failure can lead to more productive outcomes.

"If it's not any good, no one will read it, and if it's good, people will read it, and so it's like a self-capping downside."

• The fear of social embarrassment is often disproportionately high compared to the actual impact it has.
• People tend to overestimate the long-term consequences of embarrassing moments.
• This fear can prevent individuals from taking actions that are rationally beneficial.

"The perceived expected value of even a small amount of embarrassment is just like so scary to people that it prevents them from doing things that are in all rationality zero downside all upside."

• The fear of embarrassment is irrationally high and inhibits rational decision-making.

"Nobody's going to care like nobody's going to give like you do some embarrassing thing you're like you know for for years when you're going to sleep oh that thing that was embarrassing and nobody else has thought about this situation ever again nobody else cares."

• People overestimate how much others remember or care about their embarrassing moments.

• A deep-seated need to be liked can drive behavior and decision-making.
• This need can be both a motivator and a source of vulnerability in interactions.

"I am dysfunctionally conscientious. I am deeply driven like the monster that drives me is a deep-seated need to be liked by people."

• The need for approval can significantly influence one's actions and decisions.

"I still find myself getting stuck in situations where I am easy to manipulate for that reason I guess I wear my heart on my sleeve."

• Being driven by the need for approval can make one more susceptible to manipulation.

• Being authentic in job applications can be more beneficial than trying to fit a mold.
• Presenting oneself honestly can lead to being hired for who you are rather than who you think the employer wants.

"When you apply for a job that you really want like your dream job the thing that people default to do is I want to show them how well I fit exactly what they want but what if you don't fit exactly what they want."

• Trying to fit a specific mold may not always be the best approach in job applications.

"If it does work they'll hire you for you not for who you know for the person they want you to be."

• Authenticity can lead to more meaningful and fitting employment opportunities.

• Systems thinking involves considering second-order outcomes and long-term impacts.
• Multi-ply thinking is complex and requires coordination but can lead to more accurate conclusions.

"I call systems thinking as almost like defined by its multiply thinking."

• Systems thinking involves considering multiple layers and outcomes.

"If you get any of the early replies wrong your entire conclusion is incorrect."

• The complexity of multi-ply thinking requires accuracy in initial assumptions.

• Compounding returns can have a significant impact over time compared to linear returns.
• Focusing on elements with compounding returns can lead to more substantial long-term benefits.

"If you have one thing that has linear returns and one thing that has compounding returns it kind of doesn't matter what like oh what's the rate of growth a linear thing it doesn't matter y'all like on a long time horizon if this one truly is compounding and is also viable this one will win."

• Compounding returns will outpace linear returns over a long time horizon.

• Large organizations often focus on detailed projections that may not reflect fundamental realities.
• Allowing teams to work on smaller, incremental improvements can lead to significant overall growth.

"People will sit there looking at these spreadsheets and they'll say well we got a spreadsheet that tells us we're going to expect 20 million daily active user growth next month and they'll sit there debating hey this field right here it says 20.8 should that be 20.9."

• Detailed projections can often be a comfort blanket rather than reflecting actual outcomes.

"Turns out if you take a thing and you make it work better and the work like one way of looking at product debt is how different is how your product actually operates from how your user thinks it operates."

• Improving product consistency and reducing product debt can lead to significant user growth.

• Self-organization can be effective for incremental improvements but may not work for medium-sized projects.
• Centralized planning is often overestimated in its effectiveness compared to self-organized efforts.

"Self-organizing stuff is really good at hill climbing on that thing lots of different people trying out different things."

• Self-organization can lead to effective incremental improvements.

"I happen to believe that the vast majority of medium-range projects that people do are complete and will never actually work anyway."

• Medium-range projects often fail, making self-organization a potentially better approach.

• Structural issues in organizations often cannot be resolved through linear efforts.
• Identifying and nurturing elements with network effects is crucial for long-term viability.

"If you have a thing that has linear or sublinear returns or fundamentally some kind of dying thing doesn't really matter what kinds of linear heroics you do the fundamentals don't change of that thing."

• Linear efforts cannot change the fundamental issues of a declining system.

"Finding a thing that has the network effect kind of capabilities possibilities and is viable to start is extremely challenging to do."

• Identifying elements with network effects is crucial for sustainable growth.

• Building a successful product is like nurturing a bonfire; it requires small, incremental growth.
• Large, planned efforts often fail to create self-sustaining systems.

"A working product a working ecosystem is like a roaring bonfire and so you want to make a new bonfire and so what you do is if you have all of resources you get the best wood and you stack it up real tall and you know and then you take your flamethrower and you're like wow look at all this flame it's amazing it's like no the whole question is did the wood catch is this fire self-sustaining or is it not."

• Successful products require nurturing and incremental growth rather than large initial efforts.

"It's much easier to start growing a bonfire by starting with a small thing a flame that is working and Arc and grow it incrementally and nurture this usage."

• Incremental growth and nurturing are more effective than large, planned efforts.

• The concept of the adjacent possible involves taking steps that are within reach and almost certain to work.
• Recognizing the small scope of the adjacent possible can lead to more effective and iterative progress.

"The adjacent possible is a design thinking frame and it's the set of actions that are within your reach that if you did them they would almost certainly work."

• The adjacent possible involves taking feasible steps that are likely to succeed.

"If you choose consistently you can Arc to a wildly different outcome than it looks like you could at the beginning."

• Consistent, iterative actions can lead to significant and unexpected outcomes.

• Having a North Star or long-term vision can guide iterative steps and prevent getting stuck in a niche.
• Combining a long-term vision with short-term iterative actions can lead to more coherent and successful outcomes.

"What you want is a approximate Northstar like a three to five-year vision that's like two pages it shouldn't list any specific teams or any specific products no pixels on it because it's easy for a bad idea to hide behind pretty pixels."

• A long-term vision or North Star can guide iterative progress without getting bogged down in specifics.

"If you do it without the Northstar you'll just follow whatever has the strongest gradient but if you instead you say what is the thing that has the strongest gradient in the general direction of the Northstar."

• Combining a North Star with iterative steps ensures progress towards meaningful and coherent outcomes.

• The transcript covers various themes related to social dynamics, organizational behavior, systems thinking, and product development.
• Key takeaways include the importance of authenticity, the limitations of centralized planning, the benefits of self-organization, and the significance of iterative progress guided by a long-term vision.

• The speaker emphasizes the importance of management coaching and therapy for personal and professional development.
• They share advice received from a coach about allowing oneself to be recruited by other companies yearly to ensure active and intentional career choices.
• The speaker reflects on their intentional decision to stay at Google and the internal social logic of large organizations.

"I think if the world would be a better place if everybody had a therapist and everybody had a management coach."

• The speaker believes that therapy and management coaching can significantly enhance productivity and personal well-being.

"Once a year, you should allow yourself to be recruited by another company and stop once you get to the point where you realize you're wasting someone's time."

• This advice encourages proactive career management and ensures that one stays in a job intentionally rather than out of fear of the unknown.

• Large organizations tend to develop complex internal social structures that can dominate and hinder external thinking.
• The speaker compares the internal social complexity of organizations to a metabolic rate that increases over time, eventually leading to inefficiency.

"Every large organization fundamentally, the internal social logic of that organization comes to dominate."

• The internal social dynamics of large organizations can overshadow their original purpose and efficiency.

"The social complexity will absorb all the space it possibly can and so it gets to the point where it kind of gets to the efficient frontier where it's just the company is just barely producing that new value."

• As organizations grow, their internal complexities can lead to a point where they produce minimal new value.

• The speaker strongly believes in the legitimacy and transformative potential of AI, particularly large language models (LLMs).
• They discuss the idea of AI as "magical duct tape," composed of society's crystallized intuition, and its potential uses.
• The speaker highlights the importance of taking AI for granted and the competitive advantage it can provide.

"I think AI is legit. I think that this is not some flash in the pan. This is not something that will pass."

• The speaker asserts that AI, especially LLMs, represents a significant technological advancement with lasting impact.

"LLMs are like magical duct tape that is principally composed of the crystallized intuition of all of society."

• LLMs encapsulate collective human knowledge and intuition, making them highly versatile and valuable.

"Companies that take AI for granted will have an interesting advantage."

• Organizations that integrate AI into their operations will likely gain a competitive edge.

• The speaker envisions a future where AI helps humanity escape the limitations of current software paradigms.
• They advocate for a more fluid, malleable, and personalized approach to software, enabled by AI.
• The speaker discusses the need for robust security and privacy models to safely integrate third-party AI agents.

"What if instead we use this magical duct tape and we escape the box?"

• The speaker suggests that AI can help break free from the constraints of current software models.

"Technology should be about active. It should be about helping people create and use hand-tuned tools to extend their agency."

• They believe technology should empower creativity and collaboration rather than passive consumption.

"You need something that allows untrusted third-party code written by others to collaborate seamlessly in a way that's safe."

• Effective integration of AI requires secure and reliable frameworks for collaboration.

• The speaker notes that AI is already widely used by individuals, often in informal and unstructured ways.
• They highlight the importance of acknowledging AI's role in education and adapting teaching methods accordingly.

"AI is very useful to individuals, more useful to individuals than to organizations currently."

• AI's flexibility and adaptability make it particularly beneficial for personal use.

"He required his students to use ChatGPT to write their things... any factual inaccuracies in the essay is your responsibility."

• This approach encourages students to use AI responsibly while maintaining accountability for their work.

• The speaker predicts that as AI-generated content becomes more prevalent, face-to-face interactions will become more valuable for establishing trust.
• They emphasize the need for tools that facilitate genuine human connections in both physical and virtual environments.

"We will as humans place more and more emphasis on in-person and face-to-face live interactions as a way of being our grounding of where what things to trust."

• In an era of AI-generated content, personal interactions will be crucial for building trust and authenticity.

• The speaker recommends "Origin of Wealth" by Eric B. Hawker for insights into complex adaptive systems and economic innovation.
• They highlight the book's relevance to understanding business and innovation as evolutionary processes.

"One of my favorite books is 'Origin of Wealth' by Eric B. Hawker."

• This book provides a comprehensive and accessible exploration of complex systems and their application to economics and innovation.