Guide

Personal context: why AI assistants are only as good as what they know about you

Updated November 2026 · by Jerry Liu, Dayflow's founder

An AI assistant only knows what is in the conversation, so it fills gaps with the average person's situation. A short context note about your role and projects fixes most of that. A daily paper or text log covers what you actually did. On a Mac, Dayflow keeps an automatic timeline you can export as Markdown when you want an assistant to see your real week.

You ask an AI assistant to help plan your week, and it tells you to “prioritize your most important tasks” and “protect time for deep work.” You ask it to draft a status update, and it produces three tidy paragraphs about “continued progress on key initiatives” that could describe anyone’s job. The writing is fine. It just isn’t about you.

So you spend ten minutes typing background. You explain the project, the client who moved the deadline, and the fact that Tuesdays are wall-to-wall meetings. The next answer is much better. Then tomorrow, in a fresh conversation, it is gone, and you start over.

This guide is about that gap. It covers why assistants fall back on generic answers, six low-effort ways to give them better context about your work, and how to decide what you would rather not share. Most of it needs nothing more than a notebook or a text file. Dayflow, the Mac app whose team publishes this blog, comes up once near the end. Details about it come from its own published documentation for version 2.2.0, reviewed in September 2026.

The short version

Why does my AI assistant keep giving generic answers?

An assistant can only work with what is in front of it: your message, the earlier part of the conversation, and whatever it has been set up to look up. When something is missing, it does not stop and ask. It fills the gap with the most typical version of your situation. Ask about “my week” with no details, and you get advice for an average week belonging to an average person.

There is a second problem on your side. The context that matters most feels so obvious to you that you forget to say it. You know the project is late. You know your manager prefers bullet points. You know Friday afternoons are useless because of the all-hands. None of that is in the conversation unless you put it there.

The third problem is memory, and not the assistant’s. Most people do not remember their own week accurately. Ask yourself on Friday what you did on Tuesday, and you will probably recall the meeting that went badly and the task you finished, but not the two hours that went to a bug or a long email thread. If you describe your week from memory, the assistant inherits the same blind spots. If you have ever ended a day wondering where did my day go, you already know how unreliable that reconstruction is.

Some assistants offer memory or project features. They can help, but what they hold onto tends to be a few loose facts about you, not the details of what you did yesterday afternoon. Context you write and keep yourself works with any assistant and stays under your control.

1. Write a one-page context note

Open a text file or a notebook page and answer these questions in a line or two each:

Keep it short. A long note goes stale fast and buries the lines that change answers.

Illustration: “I’m a freelance designer. Current work: a checkout redesign for a small retailer, due the 24th, and a logo refresh that’s waiting on client feedback. My main client wants weekly updates in five bullets or fewer. Mondays and Thursdays I have calls until 1 p.m. I want to stop underquoting projects. Please keep answers short and skip the pep talk.”

How to tell after a week: Count how many times you had to re-explain background mid-conversation. If that number dropped, the note is doing its job. If you kept adding the same missing detail, that detail belongs in the note.

2. Keep a three-line daily log

The context note covers who you are. A daily log covers what you actually did. At the end of each day, write three lines:

  1. What I worked on.
  2. What got stuck, and why.
  3. What is next.

Paper is fine. So is a note on your phone. If you want a structure to copy, the daily work log template has one. The point is to write it while the day is still fresh, not to make it pretty.

Illustration: “Worked on: checkout flow screens 3 to 5, call with client about shipping options. Stuck: waiting on their payment provider’s field list. Next: mobile layout for screen 4, chase the field list.”

When you ask an assistant to draft an update or plan tomorrow, paste the last few entries first. You will notice the difference straight away, because the assistant is now working from a record instead of your Friday-afternoon guess. There is more on building this habit in how to remember what you worked on.

How to tell after a week: Look at how many days you skipped. Missing one or two is normal. If you missed most of them, the log is too long or too late in the day. Shrink it to one line, or move it to right after lunch plus a quick note before you close the laptop.

3. Give the assistant the situation, not just the task

Before you type a request, add one sentence about who the output is for and what they care about. Then add one sentence about what has already happened.

This takes about fifteen seconds and needs no notes at all. It is the cheapest fix on this list.

Illustration: Instead of “Write a status update on the checkout redesign,” try: “Write a status update on the checkout redesign for a client who is nervous about the launch date. We’re on track for the 24th but blocked on one field list from their payment provider. Five bullets max.”

The first version gets you a generic template. The second gets you something you could send with light edits.

How to tell after a week: Track how much you edit each answer before using it. If you are rewriting less, keep going. If you are still rewriting a lot, notice what you keep changing. That is the context you are leaving out.

4. Do a ten-minute Friday review

Once a week, read back over your daily logs and write a short summary: what moved, what did not, and what you want next week to look like. The weekly work log template gives you a simple layout.

This does two things. It turns scattered daily notes into something an assistant can use for planning, and it shows you patterns you would otherwise miss. If you want a rough sense of where your hours go before you start, the where does my week go calculator is a quick way to sketch it out.

Illustration: “This week: finished screens 3 to 5, logo refresh still waiting on feedback for the second week running. Lost most of Wednesday to a surprise revision request. Next week: mobile layouts, and send the logo client a firm deadline.”

Paste that summary into a new conversation along with your context note and ask for help planning next week. The assistant now knows about the stalled client and the surprise revisions, and it can plan around both.

How to tell after a week: After two or three Fridays, read your summaries side by side. If they show you something you did not already know, such as the same project stalling every week, the review is working. If they all say “busy week, lots of meetings,” add one specific number or example to each.

5. Decide what you will never share

More context gives better answers, but not everything belongs in a chat window. Before you build the habit, make a short “never paste” list and stick it on your monitor or at the top of your context note.

Common entries:

Then read the terms and settings of whatever assistant you use, so you know whether conversations are stored and for how long. When you paste a log or export, skim it first and cut anything on your list.

Illustration: Before pasting a week of logs, you replace the client’s company name with “Client A” and delete a line mentioning a colleague’s medical leave. The assistant’s answer is just as useful without either detail.

How to tell after a week: Check whether you have caught yourself about to paste something on the list. If you have, the list is doing its job. If you have never once paused, either your list is too short or you are not skimming before you paste.

6. Ask someone to read your context note

Give your context note to a friend or colleague and ask one question: “If you knew only this, what would you get wrong about my work?”

People who know you are good at spotting the obvious thing you forgot to write down, like the fact that you are part-time, or that one client brings in most of your income, or that you are trying to move into a different role. An assistant would get those wrong too.

Illustration: A colleague reads your note and says, “This doesn’t mention that you’re the only designer on the team, so every request lands on you.” You add one line, and the next time you ask for help prioritizing, the assistant stops suggesting you “delegate the visual work.”

How to tell after a week: If the next few answers feel noticeably more like they were written for you, the gap your reader found was a real one. If nothing changed, ask a second person.

Where seeing your actual day helps

The weak point in all of the above is that it depends on you remembering, and writing things down at the end of a long day. Some days that will not happen.

Dayflow is a Mac app that records the screen at one frame every 10 seconds and has AI write the day into a readable timeline. You do not have to start a timer or reconstruct your day from memory. Recordings and the timeline are stored on your Mac. You choose how the analysis runs: a model on your own Mac (screenshot analysis stays on-device), your own cloud provider, or Dayflow Pro, which deletes screenshots after processing under its privacy policy. The local and cloud processing page explains the differences.

For assistant context, the useful part is getting that record out. You can export a date range as Markdown from Settings → Export, and the export work history workflow shows how. The integrations page covers connecting assistants. Keep one boundary in mind: an assistant can receive the Dayflow data it retrieves in your conversation, and local storage does not make a cloud assistant conversation local.

AI summaries can also miss or misread work, so check the timeline before you share it. Dayflow is $0 and MIT licensed, with an optional Pro plan at $20 a month, and needs macOS 14 or later.

Dayflow publishes this article, so weigh this section accordingly.

Where Dayflow is the wrong pick

You are not on a Mac. Dayflow runs only on macOS 14 or later. On Windows, Linux or a phone, the paper or text-file log from step 2 is the better tool.

Most of your work happens off-screen. Dayflow sees only what is on your Mac’s screen. Phone calls, whiteboard sessions, site visits and hallway conversations do not show up. If that is most of your week, a written log captures more.

You need someone to tell you what matters. Dayflow is a record. It is not a coach, and it will not write your context note or decide your priorities. Steps 1 and 6 do that work.

Distraction is the real problem. Dayflow does not block websites. If you already know where your attention leaks and you want a hard stop, a dedicated blocker does that job better.

You need invoices, payroll or team reports. Dayflow does not provide any of these. Other Mac time trackers built around billing and teams are a better fit.

You are not comfortable with screen recording. That is a reasonable line to draw. A three-line daily log gives an assistant most of the same context, and nothing records your screen.

The bottom line

An AI assistant is only as specific as what you tell it. Left to guess, it describes an average person’s average week. A one-page context note, a short daily log and one extra sentence of situation in each request close most of that gap, and none of them need any software beyond the assistant itself.

Start with the context note today. Add the daily log tomorrow. After a week, look at how much you are editing the assistant’s answers. If you are on a Mac and the logging is the part that keeps slipping, an automatic timeline can fill that hole, as long as you stay deliberate about what leaves your machine.

Frequently asked questions

Why does my AI assistant give such generic advice?

It only knows what is in the conversation, plus whatever it is set up to retrieve. When details are missing, it fills the gap with the most typical version of your situation, which reads like advice for everyone. Adding a few specifics, such as your role, your current projects and what already happened this week, usually changes the answer more than rewording the question does.

What should I put in a personal context note for AI?

Keep it to things that change answers: your role in one sentence, the two or three projects that matter right now, who you report to or work for, your fixed weekly commitments, how you like answers formatted, and what you are trying to get better at. Leave out anything you would not want stored by the service, such as client names under contract, health details or passwords.

How long should a context note for an AI assistant be?

Short enough that you will actually keep it current. Many people find a single page, or roughly ten to fifteen lines, is plenty. A long note goes stale quickly and buries the parts that matter. If a line has not changed an answer in a month, cut it. Update the projects section whenever a project starts or ends, and reread the whole thing once a month.

Is it safe to share my work history with an AI assistant?

It depends on the assistant, its settings and its terms, so read those first. A good habit is to decide in advance what you will never paste, such as client secrets, credentials or other people's personal details, and to trim exports before sharing. Keeping a record on your own Mac does not make a cloud assistant conversation local once you send that record into it.

Do AI assistants remember context between conversations?

Some offer memory or project features and some do not, and the details vary by product and settings. Even where memory exists, what it keeps tends to be a handful of general facts rather than what you worked on yesterday afternoon. Keeping your own context note and work log means you control what the assistant sees, regardless of which assistant you use.

How do I give an AI assistant context about what I did this week?

Paste a short log of the week rather than describing it from memory. A three-line daily entry covering what you worked on, what got stuck and what is next is enough. Ask the assistant to summarize or plan from that log, then check its output against your own memory, since summaries can drop or misread details. An automatic timeline export works the same way if you have one.

Can I keep my work history on my Mac and still use AI on it?

Yes, if the AI runs on your Mac too. Dayflow's Local AI mode runs a supported model on your own hardware, so screenshot analysis stays on-device, and recordings and the timeline are stored locally. The tradeoff is that local models need your Mac's memory and processing power. If you later paste that history into a cloud assistant, that conversation is no longer local.

Is Dayflow the right tool if I am on Windows or my phone?

No. Dayflow runs only on macOS 14 or later and only sees what is on your Mac's screen. On Windows, or if most of your work happens on a phone or away from a computer, a plain daily log in a notebook or text file beats it, because it works anywhere and captures conversations and offline work that a screen record never sees.

What beats Dayflow if my real problem is distraction, not memory?

A dedicated website or app blocker. Dayflow records what happened and lets AI describe it, but it does not block websites or stop you from opening anything. If you already know where your attention goes and you need a hard stop, a blocker does that job directly. A timeline can still help afterward, to see whether the blocker changed your day.

How much does Dayflow cost?

Dayflow is $0 and MIT licensed. There is an optional Pro plan at $20 a month, in which Dayflow manages the cloud analysis for you. You can also use your own AI provider, where that provider's charges and terms apply, or run a model locally on your Mac. It requires macOS 14 or later and does not include invoicing, payroll or website blocking.

See where your day actually went

Dayflow is $0 and open source. Full recordings stay on your Mac, with local AI available for on-device analysis.

Download for MacOS macOS 14+ · MIT licensed · no account required