How founders should set weekly priorities

Priority setting for founders fails when the calendar opens with a wish list and no hard limit on active work. Set a capacity ledger, rank candidates into Commit / Park / Kill lanes, lock assignees and finish lines before the first deep-work block, and require a swap ticket for anything that arrives later. You want finished artifacts, not a busy-looking board.

Priority setting for founders starts with a WIP cap

Write the capacity number before you rank ideas. A solo operator often holds three to five committed outcomes. A two-person founding pair often holds six to eight shared outcomes when both people deliver, not when both people comment. Count Outcome tasks that must finish inside the current cycle, not every floating idea in Notes.

Treat the number as a physical limit. If five slots are full, a sixth item does not join as "also important." It waits, swaps, or dies. Teams that skip the limit rebuild the same overloaded Monday and call it hustle.

Stage changes the mix. Pre-seed days skew toward product proof. Seed days mix founder-led selling with product. Later stages add coordination. The founders day by stage guide is a useful check that your capacity mix matches the stage you are actually in, not the stage you wish you had already reached.

Put the capacity number at the top of the Monday note. Share it with anyone who can dump work on you. A hidden limit is decoration.

Common capacity mistakes to refuse on sight:

• Counting meetings as Outcomes. A meeting is input. The Outcome is the decision, artifact, or customer action that followed.
• Letting "research" occupy a slot with no dated deliverable. Research commits need a one-pager, a go or no-go, or a customer quote pack before the cycle ends.
• Splitting one Outcome across three assignees. Shared custody usually means no custody.
• Inflating capacity after a clean cycle. Raise the limit only after two clean closes in a row, then by one slot.

Build a candidate ledger before Monday deep work

Collect candidates late Sunday or early Monday: carryovers, customer promises, fundraising prep, hiring steps, ops debt, and personal maker blocks. Rank each row the same way so debate stays short.

Use four columns only:

Proof inside this cycle means an outsider can see a result before the next Monday. Reversible means you can undo without burning trust. Hours to Done is calendar time, not optimism. Lane is Commit, Park, or Kill. Keep one ledger so arguments stay on rows, not vibes.

If two candidates share the same proof and one is irreversible, prefer the reversible item unless cash or a signed customer date forces the door. For one-way doors, jot the rationale with confidence and a revisit date so chat memory is not the archive. Decision Loop can hold that short journal when you want a dated record.

Ledger tips that keep ranking honest:

• Prefer customer-visible proof over internal polish when lanes are close.
• Cap Commit rows. If everything wants Commit, the ledger failed. Force a forced rank of the top five before you fill slots.
• Timebox ranking to twenty-five minutes. Endless debate usually means the Done definition is vague, not that the company lacks strategy.
• When partners disagree on a row, write the disagreement as an assumption rather than soft-loading both items into the same cycle.

Commit slots, park lanes, and kill rules

Translate lanes into a capacity map before you touch the calendar.

Commit. These occupy slots. Each has one assignee, a Done definition for the cycle end, and a single source of truth. Prefer Outcome tasks that are pass or fail over vague "progress on X" language.

Park. These stay visible without consuming a slot. Park is not shame. It is a holding lane with a revisit date. If a parked item returns three cycles in a row without rising to Commit, Kill it or rewrite the problem.

Kill. These leave the active list. Write one line why. Killing an idea once is cheaper than rediscovering it every Monday.

Capacity example for a three-slot cycle:

If a slot lacks an assignee or a Done line, it is not committed. Demote it to Park before deep work starts.

Park and Kill hygiene:

• Scan Park every Monday before new ranking. Stale Park rows older than three cycles need Kill, rewrite, or a dated experiment.
• Kill reasons should be reusable: wrong ICP, no assignee capacity, depends on a hire you will not make this quarter, or the learning already arrived another way.
• Do not hide Kills in personal Notes. A shared Kill line stops the idea from reappearing as a "new" request mid-cycle.

Put the cycle on Task Boards without fake urgency

Move Commit items into Task Boards as Outcome tasks with due dates inside the cycle. Use KPI tasks only when the Done state is a number you will actually measure before the next Monday. Do not mark every idea Urgent. Urgent is for cash, outages, or signed customer dates.

Board hygiene that keeps the WIP honest:

• One board or a clear weekly list for this cycle's Commits, not five overlapping boards.
• Assignees match the slot names in your capacity map.
• Discussion stays on the task thread so the next person can see blockers without hunting chat history.
• Attachments link to the artifact, not a folder maze.

The Chief of Staff Briefing can surface blockers and pipeline the next morning. It only helps if module inputs and task states are current. Stale boards produce pretty ranks of yesterday's fiction.

Mid-cycle adds need a swap ticket

New requests will arrive. Treat each as a trade against the capacity ledger.

Ask three questions in writing on a swap ticket:

1. What Commit dies or parks if this joins?
2. What proof appears before the next Monday if we take it?
3. Who is the assignee, with a name and hours?

If nobody can answer, the request waits until the next ranking session. Emergency exceptions are rare: production down, payroll risk, or a customer contract with a dated obligation. "A partner emailed" is not an emergency by itself.

When you accept a swap, update the capacity map and the Task Boards row the same hour. Leaving the old Commit half-alive creates silent overload. Say the kill out loud to the person who held the displaced slot.

Avoid turning Tuesday into a second planning day. Ten minutes to process swap tickets is enough. Longer than that usually means the original ranking was soft.

Example of a swap done well: a prospect asks for a custom demo deck on Wednesday. Slot 2 was five discovery calls. You park two remaining calls, accept the deck with a Thursday Done line, and update Task Boards the same hour. Example of a swap done poorly: you keep the calls, add the deck, and "squeeze it in at night." That is how the cycle fails and the next Monday inherits guilt.

Reset the ledger so the next Monday can choose

The cycle end needs a reset, not a status parade.

For each Commit, mark pass or fail against the Done line. Move unfinished work to Park with a reason, or Kill it. Capture one sentence of learning per fail. Carryovers without a reason become permanent board clutter.

Then clear the Commit area so the next ranking starts from an empty commit list plus the Park lane. Teams that leave last cycle's titles in place "for continuity" quietly inherit last cycle's overload.

A simple reset script:

1. Open the capacity map and Task Boards side by side.
2. Mark each Outcome pass or fail with a date stamp.
3. Move fails to Park with one cause line, or Kill if the premise died.
4. Note one process fix for the next ranking (for example, "hours estimate was half of reality").
5. Clear Commit slots so Monday starts empty.

If you use Nautis as your AI business operating system, keep the same modules as systems of record cycle after cycle. Switching boards every week recreates the tool sprawl you were trying to escape.

After four consecutive clean closes, write down your real average hours per Commit type. Use that private calibration the next time someone asks you to "just add one more." Numbers beat optimism when the WIP cap is under pressure.

YC's published essential startup advice stresses focus and shipping over process theater. Your method should leave more finished artifacts, not more frameworks.

Keep the capacity ledger in the same place every cycle. Moving it between Docs, Sheets, and chat guarantees someone will plan from a stale copy. Link the ledger from the Task Boards description so assignees open one path.