What it does
Run a group through a backlog and come out the other side with a short, defensible action sequence everyone can explain, in under an hour.
Currently
Live MVP. Browser-only, no signup, boards stay in local storage. The full five-step method works, including the promote-to-solutions second board, share links, and Markdown, CSV, image, and JSON export. Live instrumentation is in; a co-edited shared board is not.
Roadmap
/now-next-later-roadmapNow/Next/Later Roadmap- First facilitated client session, to test the method under real disagreement
- Mobile pass on the grid (sliders carry it today, dragging is cramped)
- A worked example that walks itself, for first-time visitors
- Co-edited shared board, so a group can rank from their own screens
- Timeboxing cues for the facilitator (the deck's 2-minute rule)
- Capture the disagreement, not just the outcome: mark contested placements
- Import from a backlog paste that carries notes and options
- Compare two runs of the same board, before and after new evidence
- A facilitator mode that hides the room's placements until everyone has voted
- Team accounts (breaks the no-signup promise, needs a real reason)
Recently shipped
- Sep 3, 2026Shipped the method end to end: collect, pick two considerations, vertical stack-rank, horizontal slide, read the quadrants, promote the winners onto a solutions board.
How I know it's working
live data · last 30 daystwo_by_two_launchedtwo_by_two_axes_chosentwo_by_two_rank_completedtwo_by_two_board_completedHow it's built
How I built it
The method is not mine in the abstract: I wrote it at Kaiser Permanente with Andi Plantenberg, taught it to teams there, and have run it in client workshops since. It already existed as a deck, a published quick-start guide, and an agent skill. What it did not have was a version you could run live with someone.
Why the sequence is the product
Almost every 2x2 tool is a grid you drop things onto. That skips the only part that does any work.
The method is two passes, in order. First you stack-rank the whole list against one consideration, vertically, with nothing sharing a level. Forcing a strict order means every comparison has to resolve: you cannot park two items side by side and call it a tie. Then, and only then, you slide each card left or right against the second consideration, without changing its height. One new question per card, against an order the room has already agreed to.
The quadrants come last. If you show the four boxes at the start, people place items in the box they wanted, and the exercise becomes decoration.
The app enforces that order, and then gives you a way out: a free placement toggle unlocks both axes at once for people who already know where things go. Turning it on is recorded on the board, so an export says how the result was reached.
What it produces
A refreshed list, do-first at the top, grouped by quadrant, plus the share-out questions the deck ends on. If your cards carry candidate solutions on their back, the winners promote onto a second board where the solutions themselves get ranked on impact and feasibility. You pick the approach the same way you picked the problem.