Sam Altman x Garry TanNever a Better Time to Do a Startup · Reading notes by Lin Yueji (林悦己)

In 2005, Sam and the first batch of YC founders were building companies tucked away in a small building in Cambridge, with PG personally cooking them dinner. Back then, YC, startups themselves, and companies made up of young technical founders with no business people were all looked down on. By 2026, Startup School is far larger than in the early days, and starting a company has gone from a fringe experiment to a choice that more and more people can discuss openly, consider seriously and actually take

Startups are cool not only because they are exciting, ambitious or able to create wealth, but because they play a special role in the economy: they keep bringing new technology, new products and new ways of organizing into the economy, they let new entrants come in and challenge the ways established companies have already accepted, and they let people with ideas keep the agency to enter, experiment and change the status quo

AI first changes the time, headcount and specialist resources execution requires. A product that once took a startup months of work can now be finished by a coding agent in minutes, and a small team can use agents to get expertise across many fields that used to be hard to hire. The change that really matters is not that old things get done faster, but that the ceiling on the problems a startup can take on has been raised

What Sam introduces is not four benefits of startups, but four environmental conditions under which startups more easily gain a relative advantage

A good network does not start by calculating whether someone could be a co-founder or be useful to you. It starts from genuine curiosity: you find this person interesting, you want to understand what they are doing, and you are willing to help within what you can afford. Long-term interaction lets the other person see your goodwill, judgment and reliability; repeated real actions gradually form mutual trust, and when the right thing comes along, you are already able to act together

Mildly helpful does not require making big sacrifices for everyone. Help can be small, but it has to be real, specific and sustainable over time: answering a question carefully, sharing useful information, giving honest feedback, or making an introduction when you are confident about it

A loose network is a loosely connected network with no single center, no formal membership and no fixed collaboration. Two people may share only one favor, one weak tie or a mutual acquaintance, yet information, trust, opportunity and collaboration can flow along those ties. It relies not on central management but on many people jointly practicing openness, helpfulness, reliability and long-term thinking

Bay Area startup culture keeps a large number of ambitious, technically capable people with the will to act inside one dense network for a long time. They may meet through events, startups, investing or mutual friends, and a small favor today can, many years later, turn into a company co-founded together, an important collaboration or some other relationship nobody could have planned

What worries Sam is not that AI will fail to create enough material wealth, but that people will lose freedom, agency and things genuinely worth doing even as material abundance arrives. Whether the AI future is getting better cannot be measured only by how much content is produced or how many tasks are completed, but by whether people gain more agency, more control over their time, more of the things they actually want to do and more long-term fulfillment

The first "three months" is the amount of work that used to take three months in the old era. The second "three months" is a full three months of effort at today's raised standard, after AI has lifted the ceiling of what is possible. It does not ask you to mechanically generate more of the same tasks, but to do, at the new bar, what three months in this era should accomplish

Sam would tell his younger self that the pain, fear and failure of building a company are real, but that in-the-moment judgments about the consequences may be inflated by fear. He would keep the ambition, drive, capacity to act and courage to take on hard things, while being a little happier along the way, trusting that even after mistakes or failure, he could still recover, learn and keep moving forward

The taste Sam means here is not only visual, UI or stylistic taste. It is closer to a capacity for discernment and choice: picking, out of the many things you could do, the problems truly worth doing; making trade-offs between conflicting goals; setting a standard for the result; and judging whether a finished result is actually valuable

Cut against many years of experience does not mean experience has lost its value. It means that when tools, costs and the limits of capability change fast, an advantage built solely on "I have done this for many years" declines, and competitive advantage partly shifts toward "how fast I can understand and use today's capabilities"