Claire Blackshaw ships VR games and contributes to Godot’s OpenXR work. She wrote up her Develop Brighton 2026 talk as the honest ledger of shipping Augmental Puzzles in Godot, plus the community PS5/PSVR2 port she started, plus why that port ended in a brand-new renderer.

Her claim is not “Godot is bad” or “Godot is ready for AAA.” It is that Godot works, fast, and that using it early had a price:

  • Core sudoku gameplay was playable in about two days; the full menu flow took about a week.
  • The early-adopter tax came to roughly £80,000 in pure engine work a Unity pipeline would not have required — not counting the console port.
  • Godot is modular by design, so you fix the parts your game needs. “A game engine is 100 to 120 things, and the parts that annoy me are probably different from the parts that annoy you.”

Two war stories carry the argument. Frame-rate spikes in a build that was otherwise ready traced back to Godot’s Label3D creating a unique material per glyph; in VR a spike is a failed frame submission, so instead of shipping they wrote a replacement text system and later moved to GPU curve fonts once the Slug patent went public domain. Separately, Godot’s MIT core has no API for users, entitlements, leaderboards, transactional saves or sessions, so every platform plugin invents its own — a compliance mess on console. They built one unified backend layer in response.

The deeper problem was structural:

  • Godot’s file model is POSIX-shaped and memcpy-heavy on devices that are unified memory plus fast NVMe.
  • The object database is mutex-heavy and the renderer is neither bindless nor threaded.
  • Shaders pass through macro templates, SPIR-V and SPIRV-Cross until they hit console register limits and simply refuse to compile.

Her conclusion: “I wasn’t building a good PS5 engine. I was porting 1990s decisions to 2020s silicon.” Hence a third render path — bindless, lock-free command streams, explicit CPU core affinity, material-driven pipelines, and platform-native shaders written directly per platform.

The takeaway for studios is the part worth stealing: adopt Godot as a seed for your own tech stack, fork hard, keep shared history so you can still pull from main, and remember the money you “save” buying an engine is money you were always going to spend bending it to your game anyway.