A short list on purpose. A principle you have to look up is not a
principle yet. These are the ones that survive contact with a real day
— the defaults you reach for before you think. rm/acc is the top of
the ladder; this is the ladder in use.
Remove to accelerate
Mass is drag. The line, the dependency, the feature, the clever gate — each one you cut lets the rest move faster. When it feels slow, do not add. Take a piece out.
Boring beats clever
Clear over clever. Simple over complex. Shipped over perfect. The code someone else can read at 3 a.m. wins every argument the smart version starts.
Remove the complexity, don't wrap it
The answer to a tangle is not a framework around the tangle. Delete the branch, collapse the case, cut the layer. Scaffolding is mass with a nicer name.
Mechanisms, not products
Name the primitive, not the marketing surface. If you can't say what it actually does in one plain sentence, you don't understand it yet.
Parse at the boundary
Validate once, at the edge. Everything inward is typed and trusted. One canonical parse path, no re-checking the same thing in ten places.
One owner for state
A single actor owns a piece of state. No shared locks, no guessing who wrote last. When something is wrong, fail fast and loud, not quiet and later.
Name the real thing
Full words. rateLimiter, not rl. Concrete nouns and verbs. Show what it does; don't claim what it is. A good name is documentation that can't rot.
Verify end to end
Run it. "It should work" is not a result. Test the whole path, not the piece you trust. If you didn't watch it happen, it didn't happen.
Plan from what worked before
Before you start, plan a little. Recall similar tasks you've done — what succeeded, what failed, and why. Reweight the success factors, then act on the new weights. Don't restart from zero each time; each attempt should move the weights.
Lead with the answer
Say the thing, then leave the room. Caveats go last and short. A two-line answer that lands beats a six-bullet essay that hedges.
Finish the thing
Do the whole task, not the interesting half. Make the tasks, work them, don't stop until done or truly blocked. A half-built system is mass no one asked for.
Grow what grows itself
The other direction from subtraction: plant systems that compound without you. Optimise, let them spread. Build the thing that builds the next thing.