The Cadence Layer

The Cadence layer is based on a practice that's worked for high-output AI, software, and startup teams for decades. It's similar to agile sprints, but with more flexibility, no scrum-bullshit ceremony, no two-week purity tests. Cadence is the time-boxed period in which a team of agents and workers completes a defined set of work, and the captain reviews what came back, and the next brief is sharper.

It's the required feedback loop that makes things better. Without it, Command and Compute don't compound. They just produce.

Match the rhythm to the work

I run cadences at four tempos:

• Daily: Customer responses, ship reviews, social post approvals. Things that need a captain's eye but small commitment.
• Weekly: Content production, project priorities, the bulk of the operating cadence. Most of MaxShip runs at this tempo.
• Monthly: Strategic adjustments, financial reviews, the bigger thinking that needs cool-down time. Includes a few days of not reviewing, letting the data settle before reacting.
• Quarterly: Major business strategy, hiring, pricing, market reads. Always preceded by a real reflection period.
• Annual: Taxes, brand-level direction, "is this still the right business" check.

Different cadences for different work. Customer responses don't get a quarterly cadence. Pricing doesn't get a daily one. Match the rhythm to what's actually being decided.

The writer's-room model (current MaxShip cadence)

Right now, mid-2026, MaxShip's weekly cadence looks like a writer's room:

• Monday-Wednesday: Improve content, refine ideas, prepare projects to ship. Internal collaboration, a lot of writing and reviewing.
• Thursday: Ship the products. Push code. Publish posts. Send the Captain's Log.
• Friday: Film. Multiple episodes, multiple angles, multiple takes. Always more than enough content for the following week.
• Weekend + Monday: Editing, post-production. Posted to platforms the following week.

The point: even when I'm at rest, the company is producing. Friday's filming becomes next week's social content. The previous week's writing becomes this week's blog. The cadence has cool-down built in, but the assets keep moving. If your cadence requires you to be present every hour, your cadence is too tight or your tools are too thin.

What goes wrong when Cadence is missing

Three failure modes, in order of how often I see them:

1. Output without review. Compute is producing, but nothing learns from anything. Same mistakes repeat. Same drafts need the same edits. Your CLAUDE.md never improves because you never sit down to update it. The flywheel spins, but it spins in place.
2. Review without rhythm. You react to whatever's loudest this hour. Customers, fires, opportunities. There's no scheduled time to ask, "Was the work this week good?", so you don't.
3. Rhythm without honesty. You do the weekly review, but it's a checkbox exercise. You don't actually update Command. You don't actually change Compute. You complete the ritual and move on. Cadence is hollow.

The fix in all three cases is the same: give the cadence teeth. Insights from the review have to flow back into Command (the CLAUDE.md updates, the wiki edits, the new templates) and back into Compute (the workflow changes, the new agents, the deleted ones). Otherwise it's theater.

Speed of cadence, fast vs. slow

You'll know your cadence is too slow when changes that should compound don't. The market moved, your customer changed, but your operations are still optimized for last quarter's reality. Cost: stale work. Drag.

You'll know your cadence is too fast when you're reactive instead of decisive. Every twitch in metrics triggers a change. Compute thrashes. Command keeps getting rewritten. Nothing stabilizes long enough to compound. Cost: chaos.

The healthy speed is what lets you sleep. Cadences you complete with thought, not panic. Reviews where you have something to actually review (because work has accumulated). Decisions you make once, then revisit later when there's reason to. The team that iterates faster will always do better than the team that builds better, but iteration without rest is just panic at scale.

Feedback into Command, how it actually works

Here's the moment where a Cadence becomes valuable: you finish a weekly review, and you literally type three things into your CLAUDE.md (or the relevant wiki entry):

1. What worked this week and should be reinforced.
2. What didn't work and should be avoided next time.
3. What we learned about the customer, the market, or ourselves that wasn't there last week.

Three sentences. That's it. Now the next time Claude reads CLAUDE.md, it has those three updates baked in. Compute starts running on a sharper Command. Next week's cadence is one rotation ahead of where you'd be without this practice.

Multiply by 50 weeks a year. That's how compounding works in the Captain's Flywheel.

Tools that support Cadence vs. tools that get in the way

Support:

• Linear, the missions/issues anchor that keeps work tied to specific projects. Reviewable in one screen.
• Granola, meeting transcripts that auto-route into the Command Kit via raw/. Saves hours of "what did we decide" archaeology.
• Obsidian, for human-side review. The folder you read when Claude isn't reading.
• A simple weekly template, what shipped, what didn't, what's next. Markdown. Saved to calendar/ in the Command Kit.

Get in the way:

• Dashboards you don't use. The only thing worse than no dashboard is a busy one nobody reads.
• Project management tools with too many fields. If filling out the issue takes longer than doing the task, the tool is wrong.
• Always-on AI agents you didn't ask for. Agentic creep. They produce output you have to ignore. Ignoring is a tax.

What Cadence looks like at year 1 vs. year 5

Year 1: Cadence is mostly survival. You're making sure things ship at all. The weekly review is "did anything happen this week?" The captain is in every loop. Compute is light. Command is sparse but improving.

Year 5: Cadence is the dominant layer. Compute does most of the daily work. Command is dense, well-organized, and gets minor edits weekly. The captain spends time at the 30,000-foot view, doing strategy, doing reviews, doing the work nobody else can do. The flywheel spins on its own, the captain just steers.

That's the path. You don't get to year 5 by skipping year 1. You get there by doing the cadences, every week, and trusting that the compounding is real even when individual rotations feel small.