The Builder's Brief / Issue #5
Do things that don't scale
Builder's Brief | buildpurdue Newsletter
Issue 05 | 10 August 2026
Let's talk about why manual, unscalable work is the key to winning your first customers.
1. Do Things That Don't Scale
Picture two founders in a stranger's apartment, adjusting blinds for a better photo. That's Airbnb, and it's the tactic behind nearly every startup that broke out early: doing the parts of the job that never scale.
Paul Graham named the instinct in his essay "Do Things That Don't Scale." Nobody shows up automatically. Founders have to go get their first users by hand, one at a time, before the company earns the right to build systems for a hundred thousand of them.
Airbnb wasn't gaining traction in New York in the summer of 2009, so Chesky and Gebbia flew out to find out why. The listing photos were bad, shot on old camera phones, and nobody could tell what they'd be paying for. So the founders rented a camera and went door to door, photographing apartments themselves. Bookings in the city roughly doubled that month. The manual work told them something no dashboard could have. They kept shooting listings by hand for about a year before hiring photographers and building a real program around it.
Stripe ran a version of the same play. Instead of sending a signup link, Patrick and John Collison sat down next to a potential customer and wrote the integration code together, on the spot. Y Combinator now calls this the "Collison installation." It never scaled past a handful of customers at a time. What it did scale was the founders' understanding of exactly where developers got stuck, which is what eventually let them build a product nobody needed to sit next to.
The pattern holds for almost any founder who looked closely at their first ten users. The unscalable work isn't a phase you rush through on the way to the real business. For a while, it is the real business, because it's the only way to learn what you're actually building before you spend a year building the wrong thing efficiently.
This week, pick one person on your list, whether that's six emails or six hundred, and do something for them that will not scale. Set up their account yourself. Get on a call and watch them use the product live. Fix their specific problem by hand instead of writing a help doc for everyone. You'll learn more in that one hour than in another week of building alone. You can always scale later on.
Written by Nishant Nair
2. Founder Opportunities
Baylor New Venture Competition: open to student teams worldwide, and the prize pool tops $200,000 across cash and in-kind resources. Deadline: October 1.
MLH Fellowship, Fall 2026: a 12-week remote alternative to a tech internship, paired with a real project and a stipend. Deadline: August 31.
CALL-E: Your Code Is Calling: turn a project into an AI agent that makes real phone calls, low barrier for a first hackathon. Deadline: September 14, 11:45am SGT.
DevNetwork API + Cloud + AI Hackathon: online or in person at API World in Santa Clara, and judges score whether your build could become a company. Deadline: September 3.
Z Fellows: $10,000 and one week in SF with founders of billion dollar companies, built for students who are not dropping out. Deadline: rolling, cohorts roughly every other month.
3. Internship Opportunity
Perforated Backpropagation enhances neural networks by adding artificial dendrites to PyTorch models, allowing for significant accuracy gains and reduced parameter counts with minimal code changes. Early users have achieved a 75% error reduction and halved model sizes.
We are seeking a Research Collaborator to work directly with our team on a publishable project. You will gain early access to our proprietary technology and receive support throughout the research and publication process. This is a unique opportunity to contribute to deep learning at the intersection of neuroscience and real-world commercial impact.
If interested, contact rorry@perforatedai.com.
Sources
Was this useful?
Your rating shapes what we publish next.