Kill your darlings — cut what doesn't serve the game, however attached you are
GDC-L1-PROTO-000501
Statement
Be willing to cut features, systems, and content you love when they don't serve the game. Emotional attachment and sunk cost are not design arguments. The discipline to remove a beloved-but-not-working element is often what separates a focused game from a bloated one.
02
Rationale
The ideas a team is most attached to are frequently the hardest to evaluate honestly — pride of authorship and the hours already spent both argue for keeping something regardless of whether it earns its place [S-kill-your-darlings]. But a game is not the sum of its clever parts; it's a focused whole, and every element that doesn't pull its weight dilutes the ones that do (the elegance argument, DESIGN-0007, and orthogonality, SYS-0005). "Killing a darling" — a mechanic you invented, a level you polished, a system you championed — hurts precisely because it was good in isolation; the point is that good in isolation isn't the bar. Cutting it usually strengthens the whole and sharpens what remains.
03
Applies when
Scope decisions, feature reviews, and any moment a beloved element isn't working or is pulling the design out of focus. Especially during production crunch and when the game feels bloated or unfocused.
04
Does not apply / Exceptions
"Kill your darlings" is not "cut everything distinctive." A game's boldest, most personal, most load-bearing ideas are often its darlings and its reason to exist — reflexively sanding off everything you love produces a safe, characterless game (a real risk if the maxim is over-applied). The test is service to the game, not attachment: cut darlings that don't serve; fight for darlings that do. Distinguish "this is precious and not working" from "this is precious and central."
05
Implementation
Judge each element by its contribution to the whole, not by effort spent or affection felt (ignore sunk cost). Greybox and playtest to get honest evidence (PROTO-0003, PLAYTEST-0001) before defending or cutting. When cutting hurts, that's normal — save the idea in a "cut features" file for a future game rather than forcing it into this one. Prefer cutting to propping up: a mechanic that needs three other mechanics to justify it is often a darling.
06
Disagreement
The real tension is what to cut: ruthless-cutting culture (focus above all, cut freely) vs. protect-the-vision culture (a game's soul lives in its risky, beloved, distinctive parts). Both fail at the extreme — one into blandness, the other into bloat. The shared rule is "serve the game," and the judgment call is which darlings do.
07
Notes
The production/scope face of elegance (DESIGN-0007) and orthogonality (SYS-0005), and a sibling of greyboxing (PROTO-0003, which partly exists so you don't get attached too early). Confidence 4.
↔
Connected principles
S