Your agents can run in parallel. Your attention cannot. StripFlowing gives the human operator a simple queue, a clean resumption point, and a repeatable rule for deciding what to touch next.
This page explains the method. The app at the root domain is where you actually run it.
If you already get the idea, go straight to the app. If not, start with the operating rules below.
The failure mode is familiar: you launch one AI run, another finishes first, a third needs input, you jump tabs to inspect it, then spend the next ten minutes reconstructing where you were in the original task.
The real bottleneck is not model throughput. It is the human supervisor repeatedly losing context. StripFlowing is a way to externalize that context so you stop paying the same re-orientation cost over and over.
Working memory is narrow, and context switches are expensive. Once multiple runs are in flight, what matters is not reacting fast to every notification. What matters is having a stable rule for what to resume next and enough saved state to resume it without rummaging through tabs and chat logs.
Every delegated task gets a strip with only three fields: project, task, and anchor.
Every strip enters a queue. You do not pick the next task by mood, urgency theater, or notification order.
When you re-enter work, the anchor tells you exactly where to land, so resumption is immediate instead of reconstructive.
Three fields on a strip. A FIFO queue. One active review at a time. That is the whole operating system.
Every time you delegate a task to an AI agent, create one strip. Three fields only: Project (which context), Task (what the agent is doing), and Anchor (the exact place to return to, such as a chat, tab, branch, or tool window). The Anchor removes the search step from task switching.
Put the strip into a visible queue. The newest strip goes on top. You always work from the bottom: the oldest task first, regardless of which agent finished first. This is FIFO for human attention. No chasing notifications. No reacting to whichever run happens to finish first.
When you're ready, pull the bottom strip, look at the Anchor, and land exactly where you need to be — no search, no reconstruction, instant re-entry. Review the output, decide the next move, and if the task continues, put the strip back on top. If the task is done, remove it. The queue becomes your single external source of truth.
Borrowed from Air Traffic Control: shift a strip sideways in the rack to signal a blocker, a critical error, or something that needs immediate attention. Unlike a notification that disappears, a cocked strip stays visually out of alignment until you handle it.
StripFlowing is a practical synthesis, not a lab-validated framework. The logic behind it is still grounded in known constraints of attention, resumption, and external memory.
Writing a strip and placing it in the bay is a brain dump. You save state outside your head so your working memory can focus on one thing (Risko & Gilbert, 2016).
Sophie Leroy (2009) showed that switching tasks without an explicit resumption plan leaves attention fragments behind. The strip is that plan — a Ready-to-Resume signal condensed into three fields.
Writing on paper produces 25% faster task completion than digital equivalents, with significantly higher hippocampal activation (Umejima et al., University of Tokyo, 2021).
Air traffic controllers have managed dozens of aircraft in parallel for decades using physical Flight Progress Strips in a rack — even with digital radar available. Strips outperform dashboards for situational awareness (Mackay, 1999).
Start with any scrap of paper or a flashcard — that's all you need. If you want something sturdier, two 3D-printable models are available on MakerWorld.
Tilted desktop rack. Designed to live next to your monitor, always in sight, never on screen.
Compact version for working on the move, from a couch, or across multiple locations.
A precision cutter to get consistent strips from A4/letter paper in seconds.