When Paul Graham published his essay 'Do Things That Don't Scale' in July 2013, it was considered controversial. Genuinely controversial. Conventional startup wisdom said build systems first, automate early, design for scale from day one. Graham said the opposite: the founders who win in the early stage are the ones willing to do manual, labor-intensive work that no sane business would sustain long-term. That essay quietly rewired how an entire generation of founders thought about the first twelve months of building a company. While growth is crucial, Graham emphasizes that founders must actively create the conditions for it through unscalable actions. These actions, though labor-intensive, are essential for establishing a strong foundation. Unscalable efforts, such as personal customer service and direct sales, are crucial for testing and validating concepts. These actions provide immediate feedback and help shape the product to meet real user needs. And personal relationships with early users build the loyalty that becomes organic referral growth later. The Airbnb story is the clearest proof. Founders Brian Chesky and his cofounder went door-to-door in New York, photographing apartments for early hosts, manually processing and uploading images using spreadsheets. No algorithm. No automation. Just two founders with cameras doing work that couldn't possibly serve a million users. But it worked — because those early hosts felt genuinely cared for, and that feeling compounded into trust, retention, and word-of-mouth that no ad budget could have bought. Graham highlights personal customer service, direct sales, and founder-driven product development as key unscalable actions. These efforts, though time-consuming, are vital for building a loyal user base and refining the product. Manual management surfaces the inefficiencies that automated systems must eventually solve. Everything built later to automate a painful manual process is immediately useful, Shiyu, because it was designed around a real problem rather than a hypothetical one. It's also cheaper upfront — you avoid designing and building for scale before you've confirmed what actually needs to scale. Graham also flags something subtler: even unconscious unscalable efforts can work, but awareness prevents founders from discarding the patterns that are actually driving early traction. He frames startups as vectors with two dimensions — building and user acquisition — and argues that neglecting either one stalls the whole machine. Early power users, though limited in resources, offer valuable insights that shape the product for broader appeal. Engaging them personally ensures a strong foundation for future growth. That's not a phase to rush through; it's the foundation everything else is built on. Here is what ties all of this together, and it matters directly for you, Shiyu. The reason founders resist unscalable work isn't laziness — it's a misapplied instinct toward efficiency. They want to build the machine before they understand what the machine should do. Graham's answer is that the concierge phase, the door-to-door phase, the manual spreadsheet phase, is not a detour from building a real company. It is how you build one. Doing unscalable work changes the company permanently for the better. Early success requires manual, unscalable effort to recruit and genuinely delight users — and only after that foundation exists does automation have anything worth scaling.