I Used to Code 10 Hours a Day. Now I Ship 4 Apps in 5 Hours. The Daily Claude Rhythm That Made the Math Work.

Twelve months ago I was burning out trying to keep four iOS apps, a SaaS, and a YouTube channel alive solo. Today I run all of them with a five-hour daily rhythm built around Claude. A chapter excerpt from the book.

The math that did not add up

Twelve months ago I was running four iOS apps, a half-built SaaS, a YouTube channel, freelance prospecting, and trying to write a book on the side. Solo. The way I was doing it was the only way I knew: ten-hour coding days, six days a week, and a constant nagging feeling that I was falling behind on at least three of those projects at any given moment.

The math did not add up. Four shipping apps each need ongoing bug fixes, feature work, App Store review responses, and customer support. A SaaS needs backend work, frontend polish, and customer conversations. A YouTube channel needs scripts, edits, thumbnails, uploads. Freelance prospecting needs proposals. A book needs chapters.

You cannot do all of that in ten hours a day, six days a week, without something breaking. What broke for me was sleep, exercise, and the quality of every project at the margin.

I rebuilt the rhythm around Claude over the next six months. Today I work roughly five focused hours a day, five days a week, and I ship more across all of those projects than I did when I was doing ten hours a day, six days a week. This is the rhythm. Pulled from chapter 11 of my book — How to Claude — which is the chapter I tell every reader to use as a literal weekly template until they evolve their own.

The shape of the day

The day has four blocks, in this order. Each block has a purpose and a tool. The total day is roughly five hours.

Block 1 — Morning briefing (30 minutes, before coffee #2).

Open Claude Code in my main repo. Read MEMORY.md (auto-loaded), then ask: "What's most urgent across active projects, given yesterday's session logs?" Claude has access to my Obsidian sessions vault and my Pinecone memory layer, so it can pull yesterday's open threads and rank them.

Output: a punch list of three to five items, ranked. I review, I cut at least one, and I commit to the remaining three. Three is the maximum. If five are urgent, only three get done today. The other two get scheduled.

This block is not for coding. It is for choosing what to code. The single most expensive mistake of my old ten-hour-day life was that I never separated choosing from executing. Mornings would start with whichever Slack message or App Store review email was at the top of the inbox, and the rest of the day was reactive.

The 30-minute morning briefing is what bought me back the agency. By 9:30 I have three things written down. The rest of the day is in service of those three things.

Block 2 — Deep focus (120 minutes, no interruptions).

Pick the highest-priority item from the morning list. Open the relevant repo, the relevant Claude Code session, the relevant Skill (more on Skills in another article). Work for two hours, uninterrupted. Phone in another room. Slack closed. Notifications off.

The Claude side of this block is conversational. I'm briefing, reviewing diffs, asking architectural questions, pushing back when the proposal isn't quite right. The output is one feature shipped, or one bug fully fixed, or one chapter drafted. Not progress on multiple things — one thing done.

This is the block that produces 60-70% of my actual output for the week. Two hours, daily, of one-thing-at-a-time focus. The reason this works is the morning briefing made the decision for me before I opened the editor. I am not deciding what to work on while I work on it.

Block 3 — Parallel dispatch (60 minutes).

This is the block that requires Claude the most. I open three to five sub-agents in parallel, each one briefed on a small, well-bounded task: "draft the response to this Apple review feedback," "summarize this customer call transcript," "scan Malt for new freelance postings matching my keywords and produce a triage," "draft a Medium article excerpt from chapter 7," "review the open PRs and post a one-line summary of each on the team channel."

Each sub-agent has a Skill that gives it its workflow. Each sub-agent reports back when done. I bounce between them as their results come in, accepting, editing, rejecting, or asking for a follow-up pass.

The dispatch block is where the multiplier lives. In the same 60 minutes I would have spent on one of these tasks in my old life, I now move five forward. The math of "four apps in five hours" only works because the dispatch block is doing the cross-project work that my old day spent in serial mode.

Block 4 — Wrap and review (30 minutes).

Close the loop on the day. Update MEMORY.md if anything material was decided. Append any decisions to the project's memory/decisions.md. Commit the day's work with descriptive messages. Send the day's outputs to whoever needs to see them (Apple Connect, the team Slack, the YouTube CMS, the book publisher).

The wrap block is not optional. The reason a daily rhythm compounds across months is that each day ends with a clean state. The state lives in Git, in MEMORY.md, in the memory files, in the published artifacts. Tomorrow's morning briefing reads from a state that was correctly summarized last night. Skipping the wrap block is how memory rots and the morning briefing becomes useless.

The four anti-patterns

The rhythm sounds simple in retrospect. The reason it took me six months to land on is that I kept tripping over the same four patterns. Naming them is the first step to not doing them.

Anti-pattern 1: Marathon coding without dispatch. The old me would open one Claude session, start one task, and go for ten hours on it. The mistake is that one task fills the day even when it didn't need to. Parking the rest of the projects costs more than the one task gains. The dispatch block exists to prevent the marathon. If you're not actively running 3-5 things in parallel for at least an hour a day, you're leaving the multiplier on the table.

Anti-pattern 2: Choosing while executing. Mornings that start with "let me see what's urgent today" while already in the editor are mornings that drift. The choosing block must happen before the executing block, with a different posture. I literally use a different physical location — kitchen counter with a notepad for the morning briefing, desk for the deep focus block. The location switch is a cue.

Anti-pattern 3: No wrap. Days that end with the editor still open and the day's commits unpushed are days where tomorrow's morning briefing starts incoherent. The wrap block ends the day. The state is canonical. The morning briefing reads from clean state.

Anti-pattern 4: Treating Claude as an interruption tool. I see this in friends who are early to Claude. They use it like Stack Overflow — bouncing into it when they're stuck, bouncing out. That breaks both the focus block (the interruption costs context) and the dispatch block (no specialization, no Skills, no compounding). Claude is either your main pair-programmer or your background dispatch team. It is not your interrupt handler.

The weekly rhythm sits on top

The daily rhythm produces volume. The weekly rhythm produces direction. Sunday evening, I sit down for 45 minutes with Claude and run a weekly review.

The review has three questions. What shipped this week? What stalled? What is the most important thing for next week?

Claude reads the week's session logs, the week's commits, the week's MEMORY.md changes. It drafts answers to all three questions. I edit. The output is a one-page weekly review that I save in my Obsidian vault and that becomes the seed for Monday's morning briefing.

The weekly review is what keeps the daily rhythm from becoming a treadmill. Without it, I'd ship three things every day, none of them moving the business toward the goals I set last quarter. The weekly review is the part where I lift my head and notice that I've spent eleven days fixing bugs in App A and zero days on the launch I'd planned for App B.

If you only do one meta-rhythm, do the weekly review. It is more valuable than the daily morning briefing. The morning briefing is execution. The weekly review is strategy.

What you'll notice in the first month

If you adopt this rhythm cold, three things will happen in the first month.

You will, in week one, feel like you're not doing enough. Five hours of focused work feels short when you're used to ten hours of unfocused work. Resist the urge to extend the day. The five-hour rhythm produces more output than the ten-hour rhythm, because each hour is fully load-bearing. Sit with the discomfort of the five-hour day for the first two weeks. Then look at what shipped.

You will, in week two, hit the dispatch block awkwardly. Running 3-5 sub-agents in parallel feels chaotic at first. Each Skill needs polish. Each handoff feels improvised. By the end of week two the rhythm is smoother and the multiplier starts to show.

You will, in week three or four, have the first weekly review that points at something genuinely surprising. A project you thought was on track that has been quietly losing ground. A project you thought was stalled that's actually further than you realized. The weekly review compounds because the meta-pattern is what the daily volume is in service of.

Pulling this together

The five-hour daily rhythm — morning briefing, deep focus, parallel dispatch, wrap — sits on top of a weekly review. Each block has a purpose. Each block uses Claude differently. The whole thing is the operating system that lets a solo dev run more projects than the math should allow.

This is chapter 11 of the book. Part 2 of the book — the onboarding part — ends with this chapter. Part 3 is about building the specialist team that the dispatch block depends on. Part 4 is about turning the rhythm into a business. The full thing is on Amazon: How to Claude — The Playbook for a One-Person Business, Kindle / paperback / hardcover.

Or just try the morning briefing tomorrow. Thirty minutes, one decision: three things to ship today. The rest of the rhythm builds itself once that decision is non-negotiable.

Andy Garcia — four iOS apps, one LegalTech SaaS, one YouTube channel, a book, all on a five-hour daily rhythm built around Claude. Full playbook in the book.