Learning by Shipping
You do not learn in order to build. You want to build, and the learning comes along because the thing demands it.
The Thesis
The cleanest statement is in Want It Bad Enough: personal projects start from wanting to do something, which creates the pull to learn whatever is needed to make it real. The learning is a means, not the end. You do not learn to build, you want to build and learn in order to do it. Source: Want It Bad Enough, 2026-04-29 Skill is a byproduct of shipping, not a prerequisite for it.
This inverts the standard order. The standard order says study, then qualify, then build. Kevin's order says want it, start building, and absorb the domain under the pressure of needing it to work.
Paul Graham's shipping-rate axiom is the company-scale version: across YC batch companies through public companies with billions in revenue, he says the best predictor of success is the rate of shipping new stuff. The wiki should treat shipping cadence as both a learning mechanism and an operating metric: a fast shipping loop produces more contact with reality, and more contact with reality teaches the team what to build next. Source: X/@paulg, 2026-06-25
Why Shipping Teaches Better
Experience over Discourse argues that the highest-fidelity signal is first-person: using the thing, building with it, talking to people who run it. Source: Experience over Discourse, 2026-04-18 Shipping forces that contact with reality. A project that has to actually work cannot be satisfied with a plausible-sounding mental model, which is exactly where reading and discourse stop.
Forward Deployed names the operating-mode version of this: the skill is not any single domain, it is the curiosity to download the next domain fast enough to be useful. Billing, security, infrastructure, design, all learned because they needed doing and there was nobody else. Source: Forward Deployed, 2026-05-09 The 30-plus projects across AI, web, games, health, and hardware are not a portfolio. They are a transcript of things learned by needing to ship them. Source: Kevin Liu
The Difficulty Engine
Induced Hurdles sharpens the pattern. Each ambition selects for a harder problem than the last, and the hurdle is the point, because that is how capability compounds. Source: Induced Hurdles, 2026-05-31 Shipping hard things is the curriculum; the difficulty is self-assigned.
Connections
| Connection | Implication |
|---|---|
| Want It Bad Enough | The wanting is what makes the learning effortless instead of forced. |
| Generalist vs Specialist | Learning by shipping is how a generalist accumulates range without formal study in each field. |
| Builder Ethos | Search before building, then learn by building the complete version of the right thing. |
Anti-Patterns
- The preparation treadmill: waiting to feel ready instead of starting and learning under load.
- Studying a domain in the abstract when a small shipped project would teach more in a week.
- Collecting tutorials as a substitute for the one real thing that forces the knowledge to stick.
Timeline
- 2026-05-31 | Compiled from Want It Bad Enough (learning is a means, not the end), Experience over Discourse (first-person signal beats discourse), and Forward Deployed (curiosity to download domains fast), grounded in the 30-plus shipped projects on Kevin's person page. Source: compiled from wiki graph
- 2026-06-15 | Referenced in Projects Before Companies — PG's "projects with friends before companies" extends learning-by-shipping: the curriculum is a cool thing you wanted to exist, not a startup idea brainstorm. Source: User, 2026-06-15
- 2026-07-03 | Added PG's shipping-rate axiom as the company-scale operating metric version of learning by shipping. Source: X/@paulg, 2026-06-25