Can we finally use C++ Modules in 2026?

The promise

Maybe. Kinda. It’s complicated. A recent write-up circulating on Lobsters and a developer’s blog walks through the same roller coaster many C++ devs know all too well: modules can brutally speed up builds in simple cases, but they’re fragile in real projects. The poster reports success importing VulkanHpp as a module in a renderer, swapping a nine‑second header parse for a near‑instant import — a clear win. And yes, modern CMake helps: recent CMake 4.x (and defaults arriving in 3.28) make the basic wiring much less witchcraft than it used to be.

The pratfalls

Where things fall apart is less about the standard and more about ecosystems and tooling. The author focuses on MSVC/Windows and documents a litany of brittle edge cases: upstream changes that suddenly break a previously working module, recursive import limits, and the dreaded IDE mismatch where code compiles but the editor’s IntelliSense trips over everything — dubbed “IntelliNonSense.” It has been reported that Microsoft claimed modules were working internally as far back as 2019; allegedly the Edge team had internal builds, but that doesn’t mean everyday projects are smooth sailing. “On my machine it works” is a refrain that cuts both ways.

The verdict

So should you rip out your #includes and start importing everything? Not yet for most teams. For targeted wins — big, painful headers like VulkanHpp — modules are worth trying. For general library consumption, expect bumps: toolchain fixes, IDE support, and upstream project cooperation are still catching up. The post’s emotional throughline is familiar: hope, a burst of joy when something finally compiles, then the slow burn of regressions. Sound familiar? The landscape is improving, but cautious optimism is the order of the day.

Sources: mropert.github.io, Lobsters