Luma is Brainloom's second game. It shipped in September 2025 to strong reviews, a 4.9 star rating, and a day-7 retention number we're proud of. None of that was guaranteed. There's a version of this story where Luma never shipped.
This is an honest account of what happened between the first commit and the launch day.
What we got right: the constraint
Before we wrote a line of code, we agreed on one constraint: Luma would only use words a literate adult would recognise. No obscure Scrabble dictionary words, no specialist vocabulary. This constraint shaped every other decision we made.
It forced us to build a better dictionary (see our previous post on this), a better hint system, and a better difficulty curve. Constraints are design tools. We don't use them enough.
What we got right: the multiplayer pivot
Luma was originally a single-player game. Four months into development, we played an early version with the whole team and realised something was missing. The game was enjoyable but not social. Word games want to be shared.
We added asynchronous multiplayer — one puzzle, two players, compare scores — in the final two months. It added significant scope. It was the right call. Multiplayer is now the most-used feature in the game.
What nearly sank us: the dictionary rewrite
Three months before launch, we discovered our word validation logic had a critical flaw. About 8% of valid words were being rejected as invalid. Players in beta were encountering this multiple times per session.
We rewrote the validation layer in two weeks. It delayed our launch by three weeks. We shipped a better game because of it. But those three weeks were the most stressful of the project, and we came close to shipping the broken version to meet an artificial deadline.
What nearly sank us: scope creep on Luma Pro
We spent six weeks building a "Pro" tier with advanced statistics, tournament mode, and a custom keyboard. Two weeks before launch, we cut all of it. The features weren't ready, the pricing model wasn't clear, and adding them would have required another two months of testing.
Cutting them was painful. We'd already announced some of them. But shipping a clean, focused game was more important than shipping a complicated one on time.
The thing we'd do differently
We'd start multiplayer earlier. It was the best feature in the game and it was almost an afterthought. The next game starts with the social layer and builds outward.