Khronos surveyed more than 400 shader developers, graphics programmers, rendering engineers, and tools builders between mid-June and mid-July 2026. Games dominate the sample at 79%, but 26% work in middleware and tools and 16% in scientific visualization. The report’s headline is not a feature request — it is a friction finding.

What actually hurts:

  • Debugging and profiling was named a top-3 problem by 53% of respondents — more than 15 points clear of the next item. Average pain scored 52/100; 40% rated it 61 or higher, and 13 people gave it the maximum 100.
  • Cross-platform porting taxes the majority: 64% spend at least some effort adapting shaders across platforms, APIs, or tools, and nearly 10% call it significant or one of their largest engineering costs.
  • Language semantics. Open responses asked for pointers, function calls, and true addressability — GPU languages that behave like systems programming languages rather than restricted DSLs.

The language and platform picture is broader than the console-graphics stereotype:

  • GLSL at 60%, HLSL at 41%, direct or tool-based SPIR-V at 39%, and Slang at 34% — fourth place, and the only entrant not shipped by a platform vendor.
  • Windows 88%, Linux 74%, macOS 35%. Khronos calls the Linux number “strikingly high,” a signal that shading work now includes HPC and AI compute.
  • Compiler stability, documentation, and overall developer experience came up repeatedly as things worth fixing in the tools that already exist.

Two things make this worth reading. The bottleneck has moved from capability to visibility — the field can ship increasingly complex shading but cannot see what its shaders are doing across vendors, and the highest-scoring standards opportunity in the survey (a cross-vendor debugging standard, 4.18/5) follows directly from that. And 73% of respondents say they would join future Khronos working groups or feedback sessions, which makes the constraint look like coordination rather than appetite.

One caveat: Khronos ran the survey, interprets it, and concludes it is well positioned to address the gaps. The numbers are concrete and self-reported; the framing is not neutral.