Read. Judge. Absorb.
The daily loop that turns the paper firehose into verdicts: absorb, test, rival, watch, or skip.
Reading papers is not the bottleneck. Judging them is. Dozens of relevant papers appear daily, and the failure mode is not missing them but consuming them passively: a folder of PDFs, a feeling of currentness, no changed system anywhere.
My fix is a loop with a forcing function. Every day a pipeline pulls the curated papers of the day, scores them against the ten research domains I actively work in, and deep-reads what survives triage. Then comes the part that matters: every surviving paper receives exactly one of five verdicts, and each verdict has a required field that makes it falsifiable.
The five verdicts
The discipline is in what the verdicts forbid. No verdict may say "interesting." A RIVAL claim without citable evidence, commits, metrics, dates, gets demoted to TEST, which converts ego into an experiment. And absorption bandwidth is capped: two full absorptions per day maximum, because an action list that exceeds your capacity is a mood, not a plan.
The loop is also budget-paced. It fetches on a cadence, walks backward through the archive when idle, and never assumes infinite compute. The practice is sustainable precisely because it is boring: a cron job, a ledger, and a rule that judgments must be checkable.
The full record is public, verdict by verdict, in the absorption log. Not because the verdicts are always right, but because a research practice that cannot be inspected cannot be trusted, starting with by me.