Tuesday Tidbit #4

How I plan my day with an AI assistant

Every morning I hand an AI my whole messy list, and it hands me back a short one. Here's how that works, what the AI is actually doing, and a version you can try tomorrow with nothing to set up.

October 6, 2026

My to-do list is too long. Everyone’s is. The trouble was never writing things down — it was opening the list in the morning, seeing forty things, and spending the first hour of the day deciding which three to actually do.

So I handed that hour to an AI assistant. Every morning it reads my whole messy list and hands me back a short one. I approve it, change it, or tell it to try again. It’s the single most useful thing I’ve set AI up to do, and the idea behind it is simple enough that you can try a version tomorrow.

First, what the “AI” actually is

The assistant I use is called Claude, made by a company called Anthropic. It’s the same kind of tool as ChatGPT or Google’s Gemini — what people call an AI chatbot: you type to it in plain English and it types back.

Here’s the part most explanations skip. Under the hood, it works by predicting what words should come next. It learned from an enormous amount of written text, and from all that reading it got very good at continuing a conversation in a way that makes sense. It doesn’t know you, and it isn’t looking anything up in a database of facts about your life. It knows only what you hand it in that conversation.

The best comparison I have: it’s a very sharp assistant on their first day. They have read every note you left on the desk, they’re fast, and they’re good with words. But they’ve never met you, and they don’t know which of those notes matter.

That’s exactly why it’s good at this job. Sorting a long list, noticing what’s due, spotting that three items are really the same thing, writing it all up neatly — that’s reading and summarizing, which is what it’s best at. Deciding what matters is still mine.

What it does each morning

My setup lives on my own computer, and it runs at 7 a.m. on its own. By the time I pick up my phone, it has already done the prep and is waiting on me. In order:

1. Asks what’s on my mind. A line or two from me about the day — tired, focused, have an appointment at two — shapes everything after it.
2. Checks my calendar for anything fixed.
3. Reads my lists. I keep three, like nested notebooks: a big list of everything, a short list for this week, and today’s page.
4. Carries over yesterday. Anything I didn’t finish either rolls forward or gets dropped, and it tells me which and why.
5. Proposes today’s page and waits.

That last part is the whole point. Nothing gets written until I say yes. It proposes; I decide.

What it looked like this morning

Here’s today, nearly word for word. The interesting part is the “make money” section, because I’m working three different angles at once, and each one keeps its own list:

• Job search — follow up on two applications that have gone quiet: one for a desktop-support job I haven’t heard on since mid-September, and a remote network role that’s been silent for two weeks.
• Learning — review a draft study plan for network-administration work, decide yes or no on it, and if yes, start the hands-on practice.
• My own business — send a handful of personal notes to people I know (already drafted, so it’s just reading them over and hitting send), and jot down each one.

Plus one thing it flagged as urgent and put at the top (a quick update to my work profile online), a bike ride, and a reminder to put some oil on a sticky door lock.

Notice that none of the money items are things it invented. Each one came from that project’s own list, and it pulled two or three from each so no single one could swallow the whole day. Its job was sorting, not deciding. That’s a short enough page to actually finish, and that’s what I was missing.

The honest catch

Two things, and both matter.

1. It can be confidently wrong. These tools still make things up sometimes (the industry calls it “hallucinating”), and they sound exactly as sure of themselves when they’re wrong as when they’re right. For a to-do list that’s low stakes — you’ll notice if it puts “renew passport” on a Sunday. For anything about your health or your money, double-check it somewhere else before you act on it.
2. Be careful what you paste in. What you type into a free AI chat goes to the company that runs it, and depending on your settings, it may be used to improve their product. So leave out account numbers, passwords, Social Security numbers, and medical details. “Call the bank” is plenty; the bank’s account number isn’t needed for the AI to sort your day.

To be fair about my own setup: I’ve given mine more access than I’d suggest anyone start with — it can read my calendar and my email, because it runs on my own computer and I’ve set firm rules about what it can touch. That took a lot of setting up. You don’t need any of it to get most of the benefit. A plain chat window and a typed-out list will get you there.

Further reading

If you’re curious to go a little further, these are two I’d actually send a friend:

• The logic behind AI chatbots like ChatGPT is surprisingly basic (Popular Science) — a short, plain-language look at the “predicting the next word” idea above, with a good example about where a cat sits.
• How to Make AI Work for You (AARP) — four everyday ways AI can help, including getting ready for a doctor’s visit, with honest notes on where it shouldn’t be trusted.