Screen recordings that look perfectly sharp on the creator's machine frequently arrive blurred and fringed after upload. The cause is in how video compression treats the kind of image an interface produces.
Video codecs are built for photographic content
Compression assumes an image made of gradual gradients and moving natural detail, and it discards information the eye is least likely to miss in that kind of picture.
An interface is the opposite: flat blocks of colour separated by hard, high-contrast edges, with fine text sitting directly on those transitions.
The assumptions that make compression efficient on photographic footage make it inefficient exactly where a screen recording carries its information.
Colour is stored at reduced resolution
Most delivery formats keep full detail for brightness and store colour at a fraction of that resolution, because human vision is far more sensitive to one than the other.
Coloured text on a coloured background loses its edge definition under that scheme, which is why red or blue interface text smears more than black on white.
This is also why syntax-highlighted code is among the worst possible content for uploaded video, since almost all of its meaning is carried in colour.
Bit allocation follows motion, not importance
An encoder spends its available data where the picture changes. A mostly static screen with a moving cursor receives very little, and detail is coarsened across the frame.
Scrolling makes this worse rather than better, because a scroll changes every pixel at once and forces the encoder to describe a whole new image on a tight budget.
Sudden full-frame changes are where blocking becomes visible, which is why text looks worst during a scroll and settles once movement stops.
Resolution mismatch adds a second pass of damage
Capturing at one resolution and delivering at another forces a resample, and interface text does not survive scaling cleanly because its detail sits at the pixel level.
Recording at the delivery resolution avoids that entirely, which is the single most effective change most creators can make.
Where scaling is unavoidable, integer ratios cause far less damage than arbitrary ones, because each output pixel maps to a whole number of input pixels.
Working with the encoder instead of against it
Enlarging interface elements before recording puts text detail well above the level compression discards, which is why competent tutorials use oversized fonts.
Reducing on-screen motion helps for the same reason: fewer changed pixels means more of the data budget spent on the text that matters.
Uploading at a higher resolution than the intended viewing size also helps, since the platform's own encode for smaller displays starts from a better source.