Early access works when it is an instrument, not a till

Baldur's Gate 3 entered early access in October 2020 as a single first act and released in full on 3 August 2023 for PC, in September for PlayStation 5 and in December for Xbox. Three years, some two and a half million participants, and a legible list of what changed along the way.

Larian added classes and subclasses, reworked the interface, rebalanced combat. But the interesting part is not numerical: the studio rewrote the companions' dialogue. Early access players complained that the characters were hostile and unpleasant to travel with, and Larian made them less prickly and more vulnerable. Feedback altered not the difficulty and not the balance, but the personalities.

Early access makes sense exactly when a studio knows in advance which question it is asking players. Do you like this person is a question. Fund another year of development is not a question.

That is the whole difference between the two kinds of early access. In the first there is a self-contained slice, a specific hypothesis and a deadline. In the second there is an unfinished game and a hope that money and time will somehow meet. The second is far more common, and it is what ruined the term's reputation.

The uncomfortable part deserves saying too. Larian could afford this: the studio had money from earlier games, it had a name, and it could go three years without shipping anything finished. A team of five paying rent has no such luxury, and telling them to do as Larian did is useless. What follows from this story is not a recipe but a test: if you cannot state what you want to learn from players, you are not opening early access, you are taking pre-orders on faith.