Why simple products beat feature-rich ones that don't work

Executive overview

Most founders build too much before getting a single user. They let an untested theory run wild instead of putting something in front of customers early and learning what's actually true.

The pattern that works: start with the smallest possible version of a product, make that one thing work reliably, and only add complexity once you have something people want.

A product with one working feature is infinitely better than a product with many features that don't work.

The trap of building before validating

• Every founder starts with a lie: that they already know what the customer wants.
• The more you build before user contact, the longer that lie compounds.
• First contact with users reveals how many of your hypotheses were actually correct.
• Boiling the ocean: trying to build a complete, fully formed product before getting a single user.
• Copying a current product (e.g., "we're the next Amazon") ignores what that product looked like at launch — Amazon started as a simple website selling only books.

OpenSea: two years of grinding before the breakthrough

• OpenSea entered YC in 2018, pivoted to NFT marketplace, and ground it out for one to two years in obscurity.
• The initial product did one thing: buy, sell, and transfer an NFT.
• They didn't build their own wallet — they let users connect third-party wallets.
• Shipping constantly (once a day or week), focused on one simple use case, they built network effects.
• The lesson: the reason it worked is they started simple, made the simple thing work, then added complexity.

Gusto: payroll only, and nothing else

• Gusto launched as Zen Payroll doing exactly one thing: run payroll online.
• Before Zen Payroll, ADP required an in-person visit and a monthly phone call just to process payroll.
• Gusto's pitch: sign up, set it up, forget about it — good payroll is invisible.
• Imagine a version that offered healthcare, benefits, 401k, and payroll, but lost your money sometimes. You would hate it.
• A hated company has many features that don't reliably work.

False validation: copying a working model from another market

• "If it works in country X, it will work in country Y" is a simple thesis that investors fund easily — and founders often fail after raising on it.
• Validation from investors, press, friends, and startup competitions does not mean the product works.
• The only validation that matters is ground truth from customers.
• No authority figure — not YC, not investors — can substitute for talking to the people with the problem.
• Common failure mode: raise a lot, hire a team of 50, ignore fundamentals for two years, then return asking what went wrong.

Revenue focus gets replaced by growth logic

• Founders often avoid the core question: does anyone pay for this?
• Common deflection: "We need to increase the number of customers who could buy before increasing those who do buy."
• Magical thinking fades when runway drops to ~12 months — the quality of decisions improves sharply under pressure.
• The best founders apply that same urgency before their back is against the wall.

Discipline over intelligence

• Many founders are mistaken for geniuses when they are simply disciplined.
• Discipline means doing the right work when no situation is forcing you to.
• For every unit of discipline, you can lose significant raw intelligence and still outperform peers who are hiring, going to conferences, and talking to reporters.
• Focused work is the differentiating variable. If you shadowed the best founders, their calendars would show more concentrated work time, nothing else.
• Discipline is trainable. Doing focused work gets less painful the more you practise it.

Hardware crowdfunding: a cautionary pattern

• The appeal: a large upfront order is cheaper to manufacture, so pre-selling via Kickstarter seems to solve the problem.
• In practice: founders typically don't know how to build the product, don't know what it should cost, don't know how long it will take, and haven't confirmed it solves a real problem.
• Taking money from strangers on the internet creates accountability — if you don't deliver, thousands of people will track you down and call you a scammer publicly.
• Best practice: don't crowdfund at the earliest stage. Do it at V2 or V3, when you've already manufactured and shipped prior versions.
• Screwing up with enough people's money can leave a permanent mark — a Wikipedia page, a reputation, a following that never goes away.
• The real lesson applies beyond hardware: crypto, pre-sales, and any model where you take money at scale before delivering carries the same risk.