Reading deeply when the machines do the skimming

For about two decades, you handled screen text by skimming. You scanned, triaged, found the one paragraph that mattered, and skipped the other forty. Skimming became a survival skill. Headlines, bullet points, and the inverted pyramid rewarded it.

That era is ending because language models now skim better than you can. Ask one for the gist of a forty-page report and you'll have it in seconds, with citations. Human skimming lost most of its economic value.

The hard part remains

When a tool absorbs the easy version of a skill, you still need the version it can't do. For reading, that means deep comprehension: holding a difficult argument in your head, following its turns, noticing weak points, and remembering it a week later. The model can summarise Notes from Underground. It cannot have read it for you.

Most "speed reading" tools still optimise for getting through text fast. They sacrifice the part you came for: finishing with the ideas still in your head. When you finish a chapter in four minutes and retain almost none of it, you didn't read faster. You forgot sooner.

Pace decides what you keep

The usual assumption is that you must choose: read fast and shallow, or slow and deep. We think that trade-off comes from how fast reading has worked. Fixed-pace rapid serial visual presentation (RSVP) flashes words at you at a constant rate, which means the rate misses most words. Easy words waste your time; hard words blow past before you've understood them.

Match the pace to the material instead, and the picture changes. Give the dense sentence the extra beat it needs. Let the throwaway clause pass. Pause where a thoughtful reader-aloud would draw breath. You keep most of the speed and recover most of the comprehension, because the reader spends time where comprehension is at risk.

Where beetread fits

beetread is a bet on this shift. It's a reader designed for the world where the machines skim and humans read deeply: fast enough to get through real books, paced carefully enough that you finish with the ideas still in your head. The next two posts explain the mechanics: why fixed-pace reading fails your memory, and how we pace every word with a language model.