I can work out the logic, but get stuck turning it into C++. How do I practice that?

Write the solution as pseudocode first, then translate one line at a time. For example, turn “count each value” into a declaration for the counter, a loop, and an update. If you can’t translate a step, that points to a specific language feature to review.

When you forget syntax, look up just that feature in a reliable C++ reference, then finish the code yourself. Repeatedly writing small pieces from memory is what makes declarations, loops, and containers feel automatic.

For a while, don’t ask AI to produce the solution. If you hit a wall, try a reference or a tiny example that demonstrates the syntax you need, then return to your own problem. Otherwise it’s easy to recognize a finished answer without getting practice constructing one.

Keep the reference lookup narrow. Searching how to declare a vector or iterate over a string is useful; pasting the whole problem and reading a complete solution is much less practice.

Make a small personal template and drill the bits you keep forgetting: function signatures, loops, indexing, maps, and frequency arrays. Write short exercises where each one uses only a couple of those pieces. You don’t need to memorize every corner of C++, just the tools you use often.

Also compile early. A compiler error about a missing bracket or wrong type is much easier to fix when the program is still small.