The solo consultant batched admin block is seven chores in one hour, and that is why it always overflows.

It is the canonical time-management playbook for one-person consulting practices. Pick a recurring window on the calendar (Friday 2:00pm to 3:30pm is the most common, Monday 8:00am the second most common), defend it, and group all back-office work into it. The pitch is that batching the chores together amortizes the cognitive cost of switching between tools so you do not let admin sprawl into your billable hours. It is the same logic that says 'do all your email at noon, never at 9am'. Almost every guide for solos prescribes the same shape: pick the window, defend it, batch.

Because the block is not one chunk of work. It is seven different chores stacked together: CRM logging, followup drafting, hours bucketing, invoicing, receivables, chasing unpaid, and the weekly retro. Each chore lives in a different app (HubSpot, Gmail, Toggl, QuickBooks, Stripe, Gmail again, Notion). Each chore has a different cognitive shape (recall vs. compose vs. arithmetic vs. judgment). The mode-switch tax inside the block alone is 15 to 25 minutes for a 90-minute window. The block was advertised to you as homogeneous and is in fact deeply heterogeneous. That is the bug.

Yes, and it does not solve the problem. The mode-switch tax scales with the number of switches, not the length of the block. A 180-minute block has roughly the same number of distinct chores as a 90-minute one, so adding 90 minutes adds maybe 30 minutes of additional throughput and 60 minutes of additional fatigue. The math we have seen on real solo-consultant calendars: doubling the block length improves chore completion by about 1.4x, not 2x. After about two hours you are running on lower-quality decisions on the work that matters most (invoicing rates, escalation rules, retro framing). The fix is not 'longer block', it is 'fewer chores in the block'.

Two. The two chores in the seven that genuinely require your judgment in the moment are: approving invoices over a meaningful threshold (your number, often $5,000), and writing the weekly retro that informs Monday 8am planning. Both are short. Both want a human in the loop. Both fit comfortably inside 30 minutes. The other five chores (CRM logging, followup drafting, hours bucketing, receivables, chasing unpaid) are deterministic enough that they can be drafted continuously through the week by Clone, then approved in a one-tap review during the same Friday 2:00pm window.

Each chore becomes an event handler in ~/.clone/memory/admin-block.md. The CRM-logging handler runs on the zoom_call_ended event, drafts a call summary into HubSpot the moment the call ends, and queues a hold if the deal is high-stakes. The followup-drafting handler runs on the same event and drafts a Gmail message in your voice using the last twelve emails you sent as the style sample. The hours-pull handler runs every Friday 1:30pm and exports Toggl into per-engagement buckets. The invoice-draft handler runs immediately after, building six invoices in QuickBooks. The receivables handler runs on the Stripe invoice_paid webhook. By 2:00pm Friday you have a review queue, not a chore list.

It is plain text, around forty lines, declared as seven named handlers with a trigger, a target app, a one-line action, and an optional hold rule. There is no graph, no UI, no zap to configure. You edit it in TextEdit. You save. The next event uses the new behavior. The full reference file in this guide shows the exact shape: handler: crm_log, handler: followup_draft, handler: hours_pull, handler: invoice_draft, handler: receivables, handler: chase_unpaid, handler: friday_retro. Each handler is between three and five lines.

No. Clone keeps the same calendar block. The shape of your Friday 2:00pm changes from 'sit down and do 14 chores' to 'sit down and review the queue Clone built'. The queue lists every action Clone took, every action it held back, every draft waiting on your approval. You see more, not less. The weekly retro that used to never happen now drafts itself on Friday 1:45pm from the actual events of the week, so when you look at it at 2:15pm you are editing not writing from scratch. The architectural principle that makes this safe is in src/components/architecture.tsx: 'Always reviewable. Every action Clone takes is logged and reversible.'

Yes, in three ways. First, Zapier zaps require per-app paid integrations and break when an app changes a UI or deprecates an event. Clone reads your screen and types, so it works with any app you can use, including the small ones with no API. Second, the configuration of a typical solo consulting Zapier stack is 5 to 12 zaps in a graph; Clone replaces that with one ~40 line markdown file you can edit in TextEdit. Third, the cost shape is different: a Zapier graph for the seven chores runs $30 to $75 a month before the first task fires, plus per-task overage. Clone is $49/month flat on the Solo plan and runs locally on your Mac.

No. Clone uses the apps you already pay for. The seven handlers each name a target app (HubSpot, Gmail, Toggl, QuickBooks, Stripe, Gmail, Notion) and Clone drives the one already open on your Mac. Switch from HubSpot to Pipedrive and you change one line in admin-block.md, not seven Zapier integrations. The architectural principle is in src/components/architecture.tsx: 'Tool agnostic by design. Clone uses the apps you already pay for.'

Most consultants on the 21-day Clone trial report two to three weeks. Week one you watch each handler fire and correct the misses (Clone drafted the followup too long, Clone used last quarter's rate card, Clone misread the meeting transcript on a name). Each correction gets written back into admin-block.md as a rule. By week two the corrections drop to one or two per day. By week three, Friday 2:00pm shows you a queue of about five holds, you approve four in a few seconds each, you write the retro in fifteen minutes, and the block is empty by 2:30pm. The block is not deleted, it is shortened from 90 minutes to 30. That is the practical end state we have observed.

It holds. Each handler in the reference admin-block.md file has an explicit hold rule. The invoice-draft handler holds anything over $5,000. The CRM-log handler holds high-stakes deals with new contacts. The chase-unpaid handler holds always (you sign off on every chase). When a hold fires, Clone stops on that one action and queues it in the review surface with the exact reason. You see it Friday 2:00pm with the rest. You approve in one tap or correct the action and approve. The default hold rules are conservative; you loosen them per client as trust builds.

Add a hold rule: 'hold all followups until friday 2:00pm'. Now the followup is still drafted 60 seconds after the call (so it captures the context while it is fresh), but it sits in the review queue rather than being sent. Friday 2:00pm you review and either send, edit and send, or kill the draft. The drafting and the sending are two different decisions. Most batching guides conflate them. Clone separates them, which is why the block can shrink without losing your judgment in the loop.