How does Garry Tan recommend founders choose what to work on?
Work on what you uniquely know and what genuinely interests you—don’t chase the headlines. First-principles knowledge and earnest obsessions beat following trends.
Video Summary
Founders should build from their unique knowledge and earnest obsessions, not chase what's currently hot.
AI and agentic systems let very small teams scale tasks that once required large organizations, but demand strong data provenance and conflict management.
Treat repeatable processes (even a markdown file) as reliable employees and iterate—first attempts will be rough.
AI can remove layers of bureaucracy, empowering people closest to the work to change the process.
Consumer AI growth depends on falling compute costs; wider adoption may be slower than Silicon Valley expects (look toward ~2027).
Work on what you uniquely know and what genuinely interests you—don’t chase the headlines. First-principles knowledge and earnest obsessions beat following trends.
Agentic systems use many automated agents to execute repeatable tasks (Garry likens a markdown file to an employee). They let tiny teams scale rapidly but require iteration, provenance tracking, and conflict resolution.
AI can remove mid-level bureaucracy, erase rigid API lines, and let workers closest to the problem change the process—turning organizational layers into codified systems rather than people-heavy roles.
High compute and distribution costs limit early consumer AI. Broader breakthroughs depend on falling costs and better tooling; Garry expects major shifts over multiple years, not instant adoption.
Build trivial projects using available models to gain practical fluency—experiment without pressure, iterate fast, and use those learnings to inform bigger product decisions.
"I wanted to work at startups, but 2003 was funny in that web 1.0 had just ended."
Garry Tan reflects on his early career in 2003, noting that the tech scene was bleak following the collapse of the Nasdaq, which left him struggling to find job opportunities in the Bay Area.
At that time, his only job offers were from Microsoft for Windows Mobile or Expedia. Although he had ambitions to start his own company, practical concerns, like student loans, pushed him towards securing stable employment.
"I was chasing what was hot instead of what was clearly the thing I knew."
Tan expresses regret about not pursuing web programming during a time when many believed it was on the decline. Instead of recognizing the potential of web development, he sought out opportunities in mobile, which he thought would be the next big thing.
He highlights the importance of staying true to one's unique interests and knowledge instead of simply following market trends, stating that a more meaningful question to ask is, "What are you interested in? What do you know uniquely?"
"Courage is actually, I think, the right thing."
In discussing founder psychology, Tan emphasizes that belief in one's own experience is crucial, especially when faced with opposing viewpoints.
He reflects on the lesson that possessing direct experience provides an edge over others and encourages founders to harness their knowledge confidently, rather than getting swayed by external opinions, especially from sources like social media.
"There's just more competition than ever."
Tan points out that the current Silicon Valley landscape is marked by intense competition, which can lead to feelings of despondency among founders.
He contrasts this with the earlier days of tech, suggesting that while the environment is challenging today, the rise of the internet has allowed individuals to find niche communities and like-minded people much more easily than before.
"That was like a $2 billion to $4 billion mistake at this point."
Tan recounts an impactful moment from his early career when he turned down an opportunity to work with Peter Thiel, believing he might get promoted at Microsoft instead.
He examines this decision not just as a personal error but as a lesson in understanding missed opportunities, reinforcing the importance of recognizing one's instincts and the potential value of taking risks.
"The future is already here. It's not evenly distributed."
Garry Tan reflects on the significance of first principles knowledge in understanding the real landscape of innovation. He emphasizes that relying purely on prevailing maps or trends can lead to missed opportunities, suggesting that true insight comes from a foundational understanding of the territory or the domain itself.
Tan cites examples of talented computer scientists in Silicon Valley identifying potential that government and established institutions often overlook. He notes that Silicon Valley possesses a unique ability to harness advancements in ways that are not accessible to other regions.
He articulates that by engaging with the actual practitioners and observing their experiences, one can uncover "secret knowledge" that offers a greater understanding of what is possible, thus creating what he describes as "gnosis."
"It's very solitary. On the day that you lose your best customer, on the day that your best engineer quits, like who you going to call?"
Tan discusses the communal aspect of Silicon Valley and how Y Combinator (YC) has transformed access for founders from diverse backgrounds. He underscores that founders need a supportive network to be candid about their challenges, which is often lacking in conventional tech events.
He identifies the significance of YC's approach to eliminate traditional barriers to entry, allowing talented people from all over to become insiders and find a sense of belonging.
The transformation from being outsiders to recognized insiders, such as achieving YC status, is highlighted as a pivotal aspect of a founder's journey within the tech ecosystem.
"Having co-founders is net really good."
Tan reflects on the changing perceptions of co-foundership, discussing how past beliefs emphasized the necessity of having co-founders for success. In modern contexts, he suggests that the landscape is shifting, enabling single founders to thrive if they can embody multiple roles.
He introduces the notion of "vibe coding" and "agentic coding," which indicates that today's founders possess a greater range of capabilities and ambitions. This shift has altered the expectations and dynamics surrounding what it means to lead a startup.
The conversation sheds light on the importance of adapting to current trends and diversifying approaches in order to innovate effectively in the rapidly evolving tech landscape.
“I had this markdown file that is basically like the perfect script on how to advise founders to increase their ambition.”
Garry Tan shares his experience of rekindling his coding passion after engaging in conversations with others who had recently started programming again.
Initially, Tan approached the coding process with curiosity, wondering what it might be like to organize informative office hours.
He recorded these sessions, which provided him with valuable data and insights into effective communication practices used by YC partners, helping him identify ways to enhance founders' ambitions.
Interestingly, he discovered that while his initial advisory scripts were comprehensive, they needed simplification, leading him to open-source a less intense version.
“Just use every model and build completely trivial things for no other reason than to have a chassis with which to learn the new technology.”
Tan emphasizes the importance of engaging with new technologies without the pressure of creating significant outcomes, suggesting that experimentation is an essential avenue for growth.
In the past, building required extensive planning, including defining roles, testing, and quality assurance. Today, with the advent of new tools, much of that burden has been alleviated, allowing for quicker execution and iteration on ideas.
The current landscape encourages creative exploration, where ideas can be tested freely rather than constrained by traditional workflows.
“It's all about practicing; it's not innate or inborn.”
The conversation touches on the concept of agency, discussing whether it is something that can be developed or if it is naturally present in a few individuals.
Tan reflects on his own journey, acknowledging that he learned from past mistakes and adjusted his approach over time, suggesting that introspection and practice lead to growth.
The discussion transitions to the notion of business loops, emphasizing the idea of creating systems where outputs can influence and improve business operations.
“A markdown file is an employee that will do the job perfectly every single time.”
Tan introduces the functionality of tools like Open Claw and how they can maximize productivity in business operations through the creation of efficient, repeatable processes.
He shares that by integrating extensive data into these agents, founders can tap into advanced capabilities that allow them to operate at an optimized level, almost akin to living in the future.
He notes that while the initial investment in such technology can be high, for leaders in a business context, the benefits in streamlined processes and productivity can be substantial.
“The first time you try to do it, it's going to be bad.”
Tan discusses the necessity of iterating on processes and systems to refine them into flawless operations.
Founders should anticipate initial failures and view them as opportunities for enhancement rather than setbacks.
By maintaining open communication with their supporting tools, founders can swiftly address issues, allowing systems to evolve into highly skilled performers capable of handling various business tasks efficiently.
This principle extends across various areas including sales, marketing, and customer support, making the practice of building, testing, and refining universally applicable in business strategy.
"Even the testing part became automated to the point where I was primarily focused on black box testing to verify functionality."
The speaker reflects on their experience in automating product management and reveals that they had developed engineering skills to ensure thorough testing, including unit and end-to-end coverage.
Despite successfully automating most processes, they ended up primarily conducting black box testing to confirm system reliability.
The discussion introduces the concept of building "agentic systems," suggesting that today's focus is on identifying bottlenecks and instructing systems to create software or documentation that alleviates those hurdles.
"We have companies now that can go from zero to fifteen million ARR in about four months with just two or three people and hundreds of agents."
The speaker discusses emerging companies that can achieve rapid growth through automation, highlighting their ability to scale operations swiftly with minimal personnel.
The notion of "recursive self-improvement" is emphasized, prompting questions about the sustainability of growth and the necessity of error correction mechanisms in such automated environments.
Concerns are raised regarding error handling and management of conflicts, suggesting that without these systems, the unchecked growth of startups might be a gamble.
"When you have a lot of data, provenance matters. Conflict management becomes vital when faced with two competing facts."
Provenance of data is crucial when dealing with large datasets as it helps in determining the correct or most recent information amidst conflicts.
The conversation touches on the need for automated systems that continuously verify data accuracy and validate that it originates from reliable sources.
Founders with real-world experience may have an advantage, as they gain insights into necessary conflict management practices developed from their history in various ventures.
"I think there is a productive, constructive conflict that at its best is the search for truth."
The speaker argues that constructive conflict is essential for finding truth within organizations, emphasizing a shift in perspective that views conflict as a positive, rather than a negative, occurrence.
The potential for systems to enable exploration of conflicting ideas while removing emotional baggage is highlighted.
This approach is noted as enabling decentralization within companies, allowing all ideas to be tested and evaluated.
"He has his agent look at meeting transcripts from all his direct reports, allowing him to identify conflicts without being present."
The conversation introduces an example of how an individual utilizes an agent to analyze meeting transcripts, gaining insights into underlying issues and conflicts within teams.
By leveraging this technology, the speaker can assess challenges and dynamics within the organization effectively, even without direct involvement in those discussions.
This capability is framed as a potential solution to common management challenges associated with growing businesses, where information typically becomes too vast for one person to manage alone.
"The time that you do spend is just so much less burdened by this stuff that you always felt like you had to wade through."
"What kind of bureaucratic hell is it where two perfectly smart people are put into a situation where there’s a fiefdom?"
"How much humanity is just wasted?"
"The whole organization layer, which a lot of product managers historically did, should not be human."
"I actually think AI erases the API line."
"The person doing the work on the line should be allowed to change the line."
"Society is way slower than you think. Government is way slower than you think."
"You're going to want something that is benevolent, knows your hopes, fears, and desires, and is consistently seeking to help you."
The future of AI will likely bring about more sophisticated tools, driven by a desire for deeper context and personalization.
People will increasingly seek AI that understands their personal aspirations, emotions, and challenges, creating a more benevolent technology landscape.
The growth of this technology is expected to be gradual; significant advancements are anticipated around the year 2027, potentially leading to "harness wars" as costs decrease and access expands.
Presently, the expense of AI technology limits its reach, but within a few years, the cost of advanced computing could drop significantly, making powerful AI accessible to a broader audience.
The dynamics of the digital landscape, likened to "browser wars," may resurface as competition intensifies when computing becomes more affordable.
"We need more consumer AI software, and in a world where most consumer products are free to try, you can't do that when you have really high costs of incremental distribution."
There is a pressing need for enhanced consumer AI applications, built on the principle of low-cost distribution.
The current high costs for AI incrementally complicate accessibility, underscoring the importance of reducing these costs for consumer products to thrive.
For the first time in decades, technology possesses emotional intelligence, yet the development of ambitious consumer products has lagged, which presents a missed opportunity.
The changing economics of AI models are poised to unlock new product categories, which could lead to a renaissance in consumer technology.
"The white pill is just believing that America still has rule of law and that people in government are trying to help."
Belief in local governance and rule of law acts as a motivational force for civic engagement and community involvement.
Personal experiences, like participating in jury duty, have deepened the understanding of the community’s structure and the role individuals play within it.
Engagement is crucial; communities, especially tech-savvy populations, must take an active role in addressing local issues, particularly regarding representation and governance.
Issues pertinent to the community, such as public education and crime prevention, warrant direct involvement from residents to achieve effective governance.
Individuals feeling connected to their local governance can better influence change and ensure that their voices contribute to shaping community policies.
"When people like us don't speak up, when we don't pay attention, and hold our media accountable, not just the government, then things get very deranged."
Civic responsibility entails advocating for community issues and ensuring media accountability, especially during crises.
As witnessed during the pandemic, public awareness and organizational efforts can foster significant dialogue around pressing social issues.
There is a moral imperative for individuals to take action against systemic injustices, even at personal risk, to protect the community and its members.
The uneven response to social issues, particularly regarding representation and public safety, calls for proactive involvement from citizens invested in their communities’ welfare.
The example of grassroots organization around educational equity demonstrates how community actions can address broader societal challenges, establishing a movement toward improvement.
"If we fix local, state, and national will fix itself."
The discussion emphasizes the idea that by addressing issues at the local level, we can subsequently influence the broader systems of state and national governance.
There is a belief that over a five to ten year period, successful individuals will emerge from a system that promotes functionality and effectiveness.
Key to this process is the importance of discerning what should be accepted and what should be rejected in order to promote progress and positive impact for the community.
The overarching goal is to cultivate a society that reflects the values and aspirations of its people.