zuhair

RSI has to be a harness because recursion is a loop, compute can be amortised by distilling behaviour into weights but the breakthrough would come from a harness. And agree, what Dream-RSI describes is a lot closer to online policy optimisation but pre-existence of a primitive doesn't have to detract from novel application.

You might get a better response with a bit more transparency. You just listed features, no theory, no insight into your stack or the unique problems you faced / solved. Maybe give people some substance to comment on.

Author here. Header-only SBO container that keeps middle insert/erase working for non-assignable element types (e.g., const members) via a rebuild-and-swap path. When it's useful: mobile constraints (zero deps), tiny-N hot paths, support for immutable types. When not: if your types are assignable or you already carry Abseil/Boost; C++26 std::inplace_vector is relevant for fixed capacity.

Repo: https://github.com/lloyal-ai/inlined-vector (MIT) vcpkg: vcpkg install lloyal-ai-inlined-vector (port merged) Benchmarks: google-benchmark harness in the repo; happy to add your results.