Claude: Writing Workflow, Pricing and Limits

Use Claude when you have material worth preserving

Claude is Anthropic’s AI assistant. For a founder or creator, one useful writing job is to turn an existing article or set of notes into several drafts while keeping the meaning intact. Give it the source, the reader and the formats you need. Then judge what changed between the source and the draft.

It’s worth evaluating when you can explain the argument and supply a short example of your voice. If your current assistant already produces acceptable work, test that workflow before adding another subscription. For editing audio or video itself, choose a media editor.

Why it’s here: Sebastian confirms that he uses and personally recommends Claude. The pilot below was run by Codex and is separate from his personal use.

Complete does not mean faithful to the source

On October 4, 2026, Codex ran three Claude writing tasks using two original AI-assisted editorial examples. Each requested one newsletter and three posts. All three responses supplied the full set within the requested lengths. One of them still introduced details that needed editing.

The source described “several competing offers.” A post in the response pictured three offers and introduced an explanation for keeping one. The post was already framed as an imagined page, so this was not a claim about a real customer. It still went beyond the task’s instruction not to add numbers or motives. Editorial review flagged those additions.

Source: several competing offers.
Draft: a page with three offers.
Editorial repair: keep the quantity unspecified, or explicitly introduce a new hypothetical example. Do not present an invented detail as something the source established.

The visible setup was Opus 5.5, Medium, Manual, with Pro shown in the account menu. Chats were fresh and outside Projects, with no retries. Memory and custom instructions were not inspected, and pasted text was handled as an attachment. These are dated writing observations, not a current model benchmark. The fixtures were AI-assisted editorial material, not Sebastian’s past writing or customer research. No human editing time was measured.

Keep a short claim ledger beside the draft

A useful review is more precise than “make it sound better.” For each specific claim, record what supports it and what the source leaves open. This is an editorial method derived from the pilot, not a separate Claude feature or a newly tested workflow.

• Claim: what the draft says, including any number or promised outcome.
• Support: the source sentence or evidence that establishes it.
• Decision: keep, narrow, label as a hypothetical example or remove.

In the offers example, the source supports competing offers but no exact count. Keeping “several” preserves the source’s scope. If you want a concrete three-offer example, introduce it as your own illustration and check that the rest of the paragraph remains hypothetical.

Use the copyable writing brief to define the audience and outputs. Add: “Preserve uncertainty. Do not add quantities, motives, outcomes or personal experience. Flag gaps.” Review voice and factual fidelity separately; a sentence can sound natural and still add an unsupported claim.

Choose the billing commitment after the trial

Claude’s official pricing, checked October 20, 2026, lists Free at $0 and Pro at $20 billed monthly or $200 paid upfront for a year. Max starts at $100/month. Usage limits apply. These dollar reference prices exclude applicable tax; check your local offer and currency. Pro does not include Claude Console API usage, which is billed separately.

The annual Pro fee equals ten $20 monthly payments. Planning example: a three-month project costs $60 on monthly Pro rather than a $200 annual commitment, assuming those monthly prices and cancellation before renewal. Annual billing becomes cheaper only if you would otherwise pay for more than ten months. That arithmetic says nothing about how many finished packs you can produce.

Our Pro pilot does not establish Free-plan capacity. The current plan page lists Projects under Pro, but our tasks ran outside Projects. A reusable Project setup needs its own check before you rely on it.

Try one pack before making it routine

Choose a source you own, remove private details and request one complete pack. Keep the original response, your claim ledger and the final edits together. Continue when you would use the result and can sustain the review. If unsupported additions keep changing the meaning, narrow the task before expanding it.

Next step: prepare the writing brief, then use the Try Claude button for your own example. For a different job, compare the writing options by task. The provider link is direct and carries no affiliate referral.