AI Guides › Claude Mastery

The Cost Of Re-Explaining Yourself Every Session

By Nigel Guy · 3 min read

Re-explaining your role, your preferences, and your current project at the top of every new chat feels like a small, harmless bit of throat-clearing — thirty seconds of typing before the real question starts. Taken one chat at a time, it barely registers. Taken across a year of daily use, it's a genuinely large amount of repeated typing spent restating facts that haven't changed since the last time you typed them, for no benefit over having stated them once.

The mechanism

1. Notice the second repetition, not the tenth. The moment you catch yourself typing a near-identical opening sentence for the second time — "I'm a [role] working on [project], and I prefer [format]" — that's the signal, not a much later point once it's become an obvious pattern.
2. Sort the repeated fact by how long it stays true. Something true for months (your role, your standing style preferences) belongs in memory or an About Me file. Something true only for this piece of work (the brief for one specific document) belongs in a Project, not repeated at the top of unrelated chats.
3. Let the built-in memory feature do the ordinary version of this automatically. It's specifically designed to pick facts like this up from normal conversation over time — see "The Memory Feature Explained" — so a lot of what you'd otherwise manually restate stops needing restating on its own, without extra setup.
4. Audit occasionally, correct once. If memory or a Project has retained something wrong, that's a two-minute fix that saves the same restating cost in reverse — you don't want to be manually correcting a stale fact every session either.
5. Measure the actual saving, not the theoretical one. If you genuinely explain the same three or four facts at the start of most chats, removing that isn't a marginal convenience — it's routinely a meaningful fraction of what you type before you even reach your actual question.

What to skip

Skip trying to eliminate every restatement — some context is genuinely one-off and re-explaining it once is simply the cost of an unusual question, not a pattern worth fixing. And skip building an elaborate manual system to track "things I've said before" — that's solving the problem with more repeated effort, when the point is to let the built-in memory and Project features carry it instead.

Guardrails

• Memory and custom instructions reduce how much you retype; they don't guarantee the resulting answer correctly applies what was retained. Check the output the same way you would if you'd just explained everything fresh.
• What memory picks up automatically, versus what needs to be stated explicitly, is a product detail that's been actively refined and will likely keep changing — check current behaviour rather than assuming last year's version of the feature.
• This is about removing genuinely repeated overhead, not about withholding context a specific question actually needs. If a particular chat is unusual enough to need extra detail, provide it — the goal is fewer unnecessary repetitions, not fewer words overall.

All 751 AI guides · JulieMango plans from £17/mo