A Three-Prompt Claude Evening Routine After Cutting Five Time Sinks

Most evenings are never actually decided. They just fill up. One two-week tracking experiment in August put a number on it: the average evening swallowed four hours and ten minutes, and almost none of it was chosen.

What filled the time was familiar. Television that nobody was really watching, since the phone was out too. A podcast that couldn't be summarized afterward. An app opened, closed and opened again inside a minute. A rewatch of something already seen. One small job postponed every night. And a good half hour in bed replaying a conversation from the afternoon.

Over a year that's more than sixty full days, roughly two months spent waiting for the day to finish instead of using it. The proof of effort was a note on a phone reading "read 20 pages", written in October, and a book opened less often than the note.

None of it was rest. It was waiting.

The audit

Advice about fixing evenings usually skips the question that matters. The trouble wasn't that the evenings were badly spent in theory. It was that nobody had decided how to spend them, and an undecided evening just drains.

The method was a bit embarrassing: a Sunday, a fortnight of notes, and every piece of the evening checked against two questions.

1. When did this last leave you feeling better than when you started, and not just relieved to be doing it?
2. Are you doing this, or just waiting for the evening to be over?

The second question is the brutal one. Most of the evening turned out to be things done because the evening was there, not because anyone wanted them. Twenty minutes later, five of six items were crossed out.

What got cancelled

The second screen. Watching something with a phone in hand meant doing neither properly, and no plot from that fortnight could be recalled. The phone now stays in another room. That single change did more than the other four combined.

The rewatching. It was called comfort, but really a show you've already seen asks nothing of you, and by nine in the evening being asked for anything feels like too much. That's a real feeling, and it doesn't need three hours.

The checking. Refreshing screens that hadn't changed since eleven minutes earlier. Counted out of curiosity on one Thursday, the phone was picked up and opened thirty-one times between nine and eleven.

The rehearsing. Lying awake at half past eleven running through something awkward from the afternoon, or something to say tomorrow. It happened every night and never once improved the actual conversation.

The postponing. Every evening carried one job that kept getting pushed back: a reply owed to someone, a quick tidy, some paperwork. None was ever large. It hung over the whole night, and the dread cost much more than doing it would have.

What stayed

• Dinner at the table, not in front of a screen
• A twenty-minute walk, if there's still daylight
• One episode of something, picked before sitting down and not while sitting
• A book by the bed, which goes better now, though not as well as hoped

Four things, all of them wanted. The rest had been filler.

Where Claude fits

You'd expect the story to be that Claude replaced the five cancelled habits. It didn't. Cutting them did that, because they weren't doing any real work, and deleting them returned about ninety minutes. The catch was a completely shapeless evening, which is how things got this way originally.

What Claude adds is a boundary at the start of the evening, much as waking up supplies one for the morning. The morning begins on schedule, ready or not. Nothing marks the start of the evening, so it drifts on until you're too worn out to pick anything, and the window for choosing is gone.

The three prompts below take about ten minutes in total. Use the first every evening, the second whenever something is clearly being avoided, and the third weekly, at the point where vague Sunday dread used to settle in. They're written to paste directly into Claude, with square brackets marking where your specifics go.

1. The ten-minute close

Run it after dinner, before sitting down. At bedtime it's too late, since the evening has already happened to you by then.

Help me close the day properly so I stop carrying it into the
evening.

Today: [WHAT ACTUALLY HAPPENED, TWO OR THREE LINES, NOT A LIST
OF ACHIEVEMENTS]
What is unfinished that is still sitting in my head: [LIST]
What I am quietly annoyed about: [ONE LINE, HONEST]
What time I want to be asleep: [TIME]

Now do four things in this order:

1. Tell me which of the unfinished things genuinely needs
   anything from me before tomorrow morning, and which ones I
   am carrying for no reason. Be blunt about the second group.

2. For anything that does need doing, tell me whether it is a
   ten minute job I should do right now, or a tomorrow job. If
   it is ten minutes, say so plainly and do not soften it.

3. Give me one sentence naming what today actually was, so it
   is finished and I am not re-running it at eleven o'clock.

4. Given the time I said I want to be asleep, tell me when the
   evening should stop. An actual time.

Do not give me an evening routine. Do not suggest journaling.
I want the day closed, not optimised.

The final instruction in step one is the one doing the work. Unprompted, a model will treat each item on your list as a job and produce an orderly plan for all of them, which only makes the evening feel heavier. Asking for bluntness about what you're hauling around for no reason is what converts a to-do list into permission to drop things.

Step four matters more than it looks. A real clock time, stated by something outside your own head, is the edge the evening lacks.

2. The avoidance check

Use this when you can feel there's something you've been walking past all evening. It takes about four minutes and has changed evenings more than the other two prompts combined.

There is one thing I have been avoiding all evening and I would
rather deal with it than carry it to bed.

The thing: [WHAT IT IS]
How long I have been avoiding it: [HONEST ANSWER]
What I think is actually putting me off: [ONE LINE]
What happens if I keep not doing it: [HONEST]

Now:

1. Tell me whether this is genuinely a five minute job that has
   grown in my head, or a real job I am right to be wary of.
   Those need completely different responses and I want to know
   which one I have.

2. If it is the five minute version, give me the first sentence,
   the first line, or the first physical step. Just the opening.
   Not a plan.

3. If it is the real version, tell me the smallest piece I could
   finish tonight so it stops being one lump.

4. Tell me honestly whether tonight is the right time, or
whether
   I should schedule it and let it go until then. Sometimes the
   answer is that it can wait, and I want to hear that when it
   is
   true.

The two categories in step one are the whole trick. Nearly everything avoided in an evening is a four-minute job in a costume, and hearing that is often enough to get up and do it. Not always, though, and a prompt that could only say "do it now" would be useless within a week. Step four keeps it honest.

3. The Sunday calibration

Ten minutes on a Sunday evening. This prompt is what displaced the low-grade dread that used to creep in around four in the afternoon.

It is Sunday evening. I want a 10-minute calibration on the week
ahead and the week behind, and then I want to stop thinking
about
it.

My evenings this week, honestly: [ONE OR TWO LINES]
What I kept meaning to do in the evening and did not: [LIST]
The worst night this week, and what was happening: [ONE LINE]
What I have already committed to for next week: [LIST]

Now write me a Sunday calibration that covers:

1. The pattern in my evenings I am repeating that I should call
   out before next week starts. Not a pep talk. The real one.

2. One thing from the list I kept meaning to do, and which
   specific evening next week it should go in, and why that one.
   One thing. Not a schedule.

3. One thing I should deliberately NOT attempt next week, even
   though I will be tempted, because it will not survive a tired
   Wednesday.

4. Two sentences to tell future me on Wednesday night when I am
   wondering why I bothered with any of this.

Assume I am tired and slightly defensive. Be kind but do not
flatter me.

That final line is no flourish. Sunday evening is when criticism lands worst, and a model that picks up on it will water everything down until it's useless. Saying how you feel up front is what gets you an honest answer that you can still stand to hear.

The Wednesday lines deserve more credit than they seem to. The person on Sunday evening and the person on Wednesday are hardly the same: the first is full of optimism and short on fatigue. A note left for the Wednesday version is the only way any of the plan lasts into midweek.

When to skip all this

It isn't for everyone, so a few honest exclusions.

• If your evenings already feel like yours and you finish the day feeling you had it, don't audit something that works.
• If what you want is a wind-down routine, look elsewhere. There's no lavender here, no screens-off rule and no ten-step ritual. Those were tried twice, and adding things didn't work.
• If the real problem is exhaustion and not indecision, no prompt helps. Four hours of fog after a hard day isn't a decision problem, and it's probably not a conversation for a chatbot.
• If you came for Claude Code or developer workflows, this covers Claude.ai, the chat interface, as used by someone who doesn't write software for a living.

Advice for starting from zero

The morning-routine industry has convinced plenty of people that the first ninety minutes decide the day. Nobody sells anything for the evening. There are no evening influencers and no evening supplement stack, which is odd, since the evening is longer, it's the part you actually own, and it's when nobody else is asking anything of you.

That's exactly why it vanishes. The morning has a start. The evening has to be given one.

In practice, Claude's role here is small: ten minutes and three prompts, one daily, one when avoiding something, one on Sunday. The aim isn't to optimize the evening. It's to decide it once, before it decides itself.

Try the first prompt tonight once you've eaten, and note how Thursday looks afterward. If only one idea sticks, make it step four of the first prompt, the request for a real time when the evening should end. That one question explains most of the improvement.