Noble Whale Studios shipped Go Up, an atmospheric incremental game, after about two years of work by two developers. Then they posted the actual numbers: 278 wishlists, 158 lifetime units, $675 gross, $517 net, 20 returns, median playtime of 1 hour 25 minutes.

The roughly $500 payout went entirely to the contracted artist and musician. Neither developer was paid.

Their framing for the whole exercise is worth stealing:

“If you want to learn to release a successful game the best thing you can do is release a successful game, the second best thing you can do is release a failed game.”

What went wrong, in their own accounting:

  • Riding the AI art wave. They built custom tooling to generate images, stitch them into scenes, generate depth maps, and composite a fake-3D parallax effect. The backlash landed anyway. Their revised lesson is sharper than “don’t use AI”: players object to AI assets consumed by the player, not to AI assistance in code or research.
  • Not building a following. Roughly 98% of effort went into the game and 2% into hyping it. Their stated target for a small indie is closer to 80/20.
  • Playtesting that stopped checking. Friends reacted strongly to the first playable build, which supplied confidence nobody re-verified as the scope grew.
  • A botched Steam Next Fest. They entered Early Access and disqualified themselves — Steam permits Early Access games and demos in Next Fest, but not released games, and buying into Early Access counts as released.
  • Channels that did nothing. Curator posts hid behind an easily missed store link, visibility rounds produced no measurable effect, and no YouTuber replied to outreach.

What went right: the game got finished. The fake-3D depth effect and the animated skill tree came out better than the team believed they were capable of, and the small community’s feedback shaped the design.

Commenters added the tactical layer the post leaves out: validate on itch.io with a free web build before committing to Steam, do not launch without roughly 1,000 wishlists going into Next Fest, keep the next project to 3–6 months, skip Early Access, and make the store short description say the genre and what the player does.

Two takeaways stand out. The AI-art distinction is operational rather than moral — visible assets in the product versus tooling behind it — and the team’s bespoke image-stitching and depth-map pipeline counted for nothing against that signal. The second is that the 98/2 split is the failure mode, not the marketing budget. Key art, GIFs, and trailers are part of making the game in players’ heads, which is the part developer-brain keeps filing as a separate job.